|
6268
|
Thalamus + AgentOverflow by @Hardcore_Real_Name_is Thalamus + AgentOverflow by @Hardcore_Real_Name_is_nitish | Stardance...
|
Hardcore_Real_Name_is_nitish
|
1
|
472.0
|
0
|
[{"ship_num": 1, "ts": "2026-0 [{"ship_num": 1, "ts": "2026-06-18T09:29:17Z", "hours": 35.0, "multiplier": null, "payout": 0, "blessing": "neutral", "text": "Reviewer Notes (School Incubator Startup) Status: Built in my school’s startup incubator; some edge-case features are undergoing rapid iteration. Auth Required: Sign up via email OTP to unlock Research/Build Modes or send >3 messages (prevents API exploitation/high costs). Credits: Runs on live premium models (Claude Opus via AWS Bedrock). Minor risk of downtime if incubator credits exhaust, but we should be fine. Thalamus AI: L4.5 Agent Platform Thalamus AI (by Aphantic Corporations) is an all-purpose Level 4.5 AI ecosystem combining multi-agent workflows, deep research, and true local QEMU OS virtualization inside your browser session via a lightweight 31MB local Node.js WebSocket bridge. The 4 Core Modes Chat: Context-aware assistant for writing, planning, and summarization. Research: Synthesizes live web search with inline source citations. Study: Adaptive RAG tutor; uploads notes to auto-generate custom quizzes/solutions. Build (Code): Multi-agent pipeline building full-stack apps from plain English. 9-Agent Code Pipeline Runs asynchronously via Convex scheduled functions using smart model routing: Researcher (Haiku): Maps existing engineering benchmarks. Analyser (Haiku): Dissects product requirements and edge cases. Planner (Haiku): Graphs detailed task execution plans. Coder (Opus 4.6/4.8): Authors core logic, APIs, and databases. Optimiser (Sonnet): Refactors code for scale and algorithmic efficiency. Organizer (Haiku): Automatically structures folder/file directories. Tester (Sonnet): Generates robust end-to-end unit test suites. Hacker (Sonnet/Opus): Conducts penetration audits to patch vulnerabilities. Critic (Haiku): Acts as the final quality gate; rejects incomplete work. Native Stack & OS Virtualization OS Support: Boots Windows 11/10, macOS Sequoia/Sonoma, Linux (Ubuntu/Kali), and Android. Browser-only instant boot for Alpine/Arch Linux. Ultra-Lean Desktop: Built with Neutralinojs (uses native Edge WebView2)—only 2.2 MB total app size. Integrated with GitHub OAuth for automated commits after agent cycles.", "repo": "https://github.com/hardcoregamingsyle/thalamus", "demo": "https://thalamus.aphantic.skinticals.com"}]...
|
[{"post_id": 37547, "ts": "202 [{"post_id": 37547, "ts": "2026-07-30T11:01:32Z", "hours": 9.0, "text": "I am now doing something kind of risky. giving AI adhd. Like literal ADHD. level 2/5 and for some tasks level 2.5/5 as per my research, this should increase the output quality of AI. also checked agentoverflow compute, it’s working Okayish. claude has access to it and my limit resets tommorow, so the automated checks aint happening and I have to check manually. Here is SS of me creating a whatsapp clone as the test of this AI(hopefully this tells you what Im expecting form my AI.)"}, {"post_id": 37363, "ts": "2026-07-30T00:55:25Z", "hours": 2.0, "text": "Everything Finally works, except the VM, it got some error, debugging that, then I just came up with a new idea(thanks to my yt feed) add a ADHD mode toggle in Chat mode and Research mode, basically ADHD in AI will I think help AI leave its very formal guided approach t problems, with the distracter, it will be jumping here and there, maybe making connections, giving us a better take at making AI good. There will also be a Brainstorm mode, the Brainstorm mode will give AI ADHD and it will help in brainstorming startup Ideas, startegies, and stuff. can someone remove this screenshot compulsion. I dont have a screenshot every devpost."}, {"post_id": 37002, "ts": "2026-07-29T13:34:59Z", "hours": 7.0, "text": "Finally, got the routing to work properly. Everyting works now. My head is so done for it. also, I just saw something, and People like me may be the reason for it getting added. Im uploading it as the screenshot."}, {"post_id": 36337, "ts": "2026-07-28T12:00:04Z", "hours": 17.0, "text": "I dont understand why is my Code mode routing going to Ollama when everyting else point to NVIDIA NIM, even the OLLAMA Fallback is Nvidia NIM, the Primary provider is convex dashboard is NIM, my ollama keys are at limits and this great code mode is pointing at NIM. whyyy. Currently working on fixing the Github sanbox. thats kindof broken. rest everything works. For some reason, even on a successful github import, it shows an error(AI slop, thats why. I made the github system with AI) . the sandbox just doesnt detect the repo. and the Hack Club Fraud Squad is sus with me(I dnt blame them, 17hrs a day for a 14yr old is suspicious.) I have migrane, I dont have glasses and my eyes are very healthy(somehow). once any 1 of my project gets reviewed and I get payout, im spending it all on hostng and AI credits so I can switch over to Oracle and Modal. Agentoverflow is still going through filteration and final review. I have cooked myself. I am still procastinating the .exe(even with AI, building a native .exe is a headache.) here is SS of the new Sandbox tab"}, {"post_id": 35769, "ts": "2026-07-27T12:28:48Z", "hours": 21.0, "text": "polishing the VM and all, and good news, Im part of the 1 million founders, 5 day bootcamp. Its a very very god program, the onluy thing i hate about it is the dress code, like seriously, shirts and pants. why not joggers and t shirts, they are confortabel af. also, i monitored teh agentoverflow corpus, well, it had a problem, the API key of gemini to verify absolute gold tier learnings, im so dumb I waited 3 days for something to happen and only to realise forgot to set the APU key. yup, this is greatness at it’s peak. for thalamus, doing some polishing, just realised that git import diesnt work, fixing that, removing the requirement of git pat. Fixing a few more stuff. also, I just realised another problem, the database. how am I going to set up databases, like how do I set uo convex for everyone who uses this for free. well if the VM problem had a solution, this should also have a solution. API is working, tested it with claude code. Claud eput in 15 learnings for me. also, I realised that I forgot to change the sandbox tab to leave the old VM system an move on to the new Github Actions system. Also, long time I havent updated my .exe gotta do that also. Quick update- I got a Notif that I am suspiciously a fraud, like a serious fraud. the Fraud Squad app sent me the notif. after that, I am still trying to recover from the emotional trauma of the database loss. I will be angry with my mother for a bit more time(se was the one who deleted it). IK im cruel and he didnt mean it, but recovery takes time. also, For the next week, after I do the changes and fixes required, I will be using only and only thalamus to code Thalamus. right, An AI that will code itself, but I will be the one guiding it. because if I cant use it, my product aint ready for market launch. for screnshot, idk what to put. like all I did was console work and Im definately not showi my console here. the names are kindof inappropriate. so here is my old sandbox tab. Good Bye QEMU and VNC Viewer. you are getting replaced by Github(atleast they aint getting replaced by AI.)"}, {"post_id": 35073, "ts": "2026-07-26T08:12:34Z", "hours": 2.0, "text": "I FINALLY CRACKED THE UNLIMITED VM PROBLEM. Github Actions are unlimited usage for publrepos, and for source control, we are already making a public repo, and we get a Free VM with it. I can Run iOS, MacOS, Windows, Linux and Android on it, no restrictions, completely free. WoHoo. also, now I am no longer using Closed source Models, because Open Sourced Models, if Used correctly are alot cheaper and provide the same result as any Closed Sourced Models. Agentoverflow is still working well. Will shift from google Cloud to Oracle Cloud, as Orace gives a 50% more powerful, compoletely dedicated server for free with 200gb storage, and that additional 150gb that I need costs me just under $4 a month, so It is very very viable financially. will switch to it later. I will now add a Student Mode for anyone with the email given by my school, and a teacher mode for teacher emails. Still recovering from the emotional trauma of the database loss. Added gravity Ads, need to remove alot of adslots because right now, it is spam impressions, not hih quality impressios, but still, here is ss."}, {"post_id": 35023, "ts": "2026-07-26T05:38:11Z", "hours": 5.0, "text": "I purged all dad code, did a massive refractor, fixed anything broken, left nothing. both repos, around 15k lines touched(must be my most productive session.) and I am so done, my migrane is at its peak, but IDK, i cant sleep anyways so better work than stare at the ceiling. Also, AdminPanel finally works. and since I lost my entiredatabase, it will take a day or 2 for the reindexing to start, right now its just botting up, downloading the embedding model and the entire database. I am still depressed. will take me some time to recover from the emotional trauma."}, {"post_id": 34718, "ts": "2026-07-25T17:33:45Z", "hours": 18.0, "text": "i am so done for it. i spent this entire week. trying to find out why is the admin panel showing 0 learnings. guess what. my mom deleted the VM, thinking it would remove all charges, I didnt even have a backup. great, back to square 1. and the agentoverflow admin panel has a error, trying to integrate AdMesh for Ad Revenue. Will use Modal as it is a serverless infrastructure for hosting AI models. I now need credits. allso, my mouse and keyboard are giving up on me. here is ss of error."}, {"post_id": 34108, "ts": "2026-07-24T16:20:33Z", "hours": 34.0, "text": "busy fixing the agentoverflow error. still working on it, no time for a detailed devlog, here is ss of admin panel, latest."}, {"post_id": 32916, "ts": "2026-07-22T11:46:35Z", "hours": 11.0, "text": "my entire databse got wiped, 0 learnings. and that isnt the worst. my greatest enemy, Daal(We caal him that), literally told my crush, who was very close to me(we talked alot), and now, she avoids me. Thanks to my friends, I know why she avoids me(I skip school almost everyday). Here is a ss of the database with 0 learnings. Im broken, depressed, you can make me happy by sending a emaill to Anthropic, telling them to integrate agentoverflow directly into their tools list for code mode."}, {"post_id": 32405, "ts": "2026-07-21T15:50:48Z", "hours": 0.0, "text": "quick Update. Now im using ollama cloud wil like 3 different accounts, will exapnd this to like 50 different accounts so I never have to go on paid tier till I have the money. I was checking hackatime, And im No. 1 here is ss for proof."}, {"post_id": 32382, "ts": "2026-07-21T14:58:46Z", "hours": 3.0, "text": "Finally, found a reliable API aource(for now). well now I can finally relax. but I wont, why, I need to fix the bug that my input tokens are at like 200k while my output tokens are barely 20k, meaning my input is bloated. so gotta fix that full night. and as I was writing this, guess what I saw.Eror 403, insufficient balance. guess im back to looking for API. god bless me with unlimited credits."}, {"post_id": 32287, "ts": "2026-07-21T11:11:58Z", "hours": 15.0, "text": "Trying to find better API for the AI Agents in Thalamus. Currently trying our Cloudflare Workers AI, in Agentoverflow, the Database Size is now 1.125Mil. I was showcasing my project in the incubator, and guess what, I was surorised to see a payment error, My API hit it’s limit somehow and I got 0 RESPONSE. I GOT HUMILIATED. for SS, I got nothing, currently working on Dynamic Compressions and stuff, als, in agentoverflow, I expect a problem in searhc, the result being not relevant. I will be sending agentoverflow to Claude Code MCP forum so the reviewers review it and put it in. here is a ss of the admin panel."}, {"post_id": 31734, "ts": "2026-07-20T11:27:54Z", "hours": 9.0, "text": "i havent slept since the last 24 hours, cause FIFA world Cup was at night, and during the day, I never sleep, so Im forcing myself to work as hell. already crossedthe 800k point. Improved the Dispatcher in Thalamus. the Thalamus test is still going on. I have started blog posting for SEO. I have indexed both sites on Google search. set both Platforms completely unrestricted and free for now. Link to both websites: https://agentoverflow.aphantic.skinticals.com/ https://thalamus.aphantic.skinticals.com/ here is SS of the AI working."}, {"post_id": 31546, "ts": "2026-07-20T01:19:26Z", "hours": 12.0, "text": "Now working o SEO and indexing, Agentoverflow db has over 730k entries. Intigrated Agentoverflow MCP into thalamus and also integrated Sketchfab MCP for 3d models in case of Gameev or projects that require a 3d model. Currently testing my Code mode. Prompt: Build a highly advanced flight sim. Also, quick question, I made this AI, so I can use my own AI to build other projects, right."}, {"post_id": 31063, "ts": "2026-07-19T04:50:28Z", "hours": 8.0, "text": "agentoverflow has 475k+ examples in dataset(total is 3.7m), and im so great, I forgot to git push and deleed my folder, great, time to redo the special feature(not releasing yet). I published my site to google analytics. Im talking with my school rn if they can take my project and make it the official school AI. my school’s beurocracy is slow as hell. so idk if it will ever happen. also, here is a ss of the admin panel(of agentoverflow)"}, {"post_id": 30698, "ts": "2026-07-18T14:21:53Z", "hours": 2.0, "text": "did the UI and animations for agentoverflow also. https://agentoverflow.aphantic.skinticals.com/"}, {"post_id": 30650, "ts": "2026-07-18T12:07:49Z", "hours": 15.0, "text": "added 3d and scroll anims to Thalamus, increased the Limits per day in Agentoverflow so it can compete with StackOverflow, currently at 300k Examples for Agentoverflow, compute is slow, will reach to 3.7m+ by 3 weeks."}, {"post_id": 30145, "ts": "2026-07-17T14:16:22Z", "hours": 4.0, "text": "Testing MCP, I am setting AI training Aside because im currently creating the database. hopefully everythng will work, this AgentOverflow is like the brilliant idea, like barely any competition. website: https://agentoverflow.aphantic.skinticals.com/ git repo: https://github.com/hardcoregamingsyle/agentoverflow currently ingesting the entire stackoverflow database, while integrating it into thalamus. will also submit agentoverflow as a tool for ClaudeCode. SS of new website."}, {"post_id": 30046, "ts": "2026-07-17T09:48:59Z", "hours": 15.0, "text": "more work on Agentoverflow, added MCP tool, currently around 10% of stackoveflow database is in this bad boy. here is a ss of what im thinking of doing next."}, {"post_id": 29402, "ts": "2026-07-16T10:20:48Z", "hours": 14.0, "text": "working on Agentoberflow, which is stackoverlow for agents, Side by side working on dynamic pipeline. For agentoverflow, as a starting point, I am adding the entire stackoverflow database, like its entire codebase, Cleaned and filtered, Vectorised, and Graphed using GraphRAG, Storing it in maybe AWS EC2 as it is the cheapest. Also working on the Agentoverflow website, though both have the same database, like the exact same database in convex so that in the future, when I want to make a My entire ecosystem at 1 place, it is very very easy for me to do. and my life is in the greatest mess possible, my parents make me constantly change setups, due to which I need my own laptop, in the sense which I got on my own with my own effort so they cant take it from me, and I also need my own phone. so I dont have to make sure that my whatsapp is logged into every device. and I am completely dropping Payments in the App because BMAC requires stripe to connect payment. here is a screenshot of the new Agentoverflow Website."}, {"post_id": 28821, "ts": "2026-07-15T10:06:51Z", "hours": 11.0, "text": "don’t ask whats going on with my self trained AI, it is in no good state, the code either works, or it doesnt, or it works but something is off. I have a new idea, under thalamus, I will have AgentOverflow AgentOverflow is like StackOverflow but for Agents, will explain it later(when I actually start to work on it) BMAC is in queue. Dynamic pipeline is kindof done, kindof undone. Desktop app turned out to be nicer than I thought. rest I have nothing to say. here is a screenshot of the desktop app build mode returning a valid output instead of an error thats all I have to say"}, {"post_id": 28252, "ts": "2026-07-14T12:22:05Z", "hours": 2.0, "text": "I was trying to put in dynamic AI coding pipeline, and guess what, My code mode compeltely crashed. My AI is still being trained(errors o errors+ hitting my weekly limit.), also,My application for Emerging Ventures got rejected because I hadnt put in all info(mb), Will reapply again. For some reason, my GravityAds Account is showing unverified Via API but verified via dashboard, and I am still working on BMAC integration for paywall, I need to figure out on the payment and stuff. this is Devpost 1 of asking for the Claude Max x20 plan in the shop. and Huggingface GPU hours credits in the shop. here is a screenshot of of my rejection letter."}, {"post_id": 27695, "ts": "2026-07-13T14:07:31Z", "hours": 10.0, "text": "gravityads yet not working, BMAC under progress, how to deploy my AI is still a question with no practical answer, I asked my parents to buy huggingface paid plan for me so I can use zeroGPU, I got a hour long lecture on How I have to escape the fake tech world, and how these people scam us. currently working on improving the feel of the desktop app and vibe coder, step-1: Give it the feel of Claude Code, Codex or any other l4 vibe coder(easy) step 2: Upgrade the feel to make it a true l4.5 Agent(near impossible level difficulty, I wont be copying anyone here, I would have to make my own definatation of L4.5) Side Quest(Very Important): Either get yourself a Good amount of GPU hours or get yourself a way to deploy your AI model somewhere. Frustration is at it’s peak. and can someone please gift me a RTX Pro 6000(Blackwell), it has 96gb of VRAM, enough to trsin a 64B Model on it and make it very very good. here is the screenshot of a response from AI in the desktop App. Now I will work on MacOS desktop app."}, {"post_id": 27099, "ts": "2026-07-12T11:22:49Z", "hours": 28.0, "text": "Adding GravityAds for some revenue(this is a startup, remember), tried to add Payment but BuyMeACoffee crashed, HCB Bank isnt in india so I had to go with BMAC, so payments is the next task, the exe had some reworks done. still looking for a way to make my $50 of AMD credits last long, also I noticed that Rather than wasting it on running, I can use the aprox 200gb of VRAM to train a very powerful AI Model in like 25 hours, and then give you all the .gguf and then it is up to you if you want to test the AI or not. also, I am now applying for Emergent Ventures(EV) and Theil Fellowship, because I need the most amount of grant possible to get this to work. and currently on a hunt to find ISOs for the QEMU VM system, I am looking for Preactivated Pro verison of Windows 10 and 11, which I think is illigal but right now, we havent launched publically yet so I dont think it matters that I am going on the illeagal path of Software Piracy, but I promise I will once use real lisenced Windows version. For MacOS, it is something that can never be on the legal route till I get a good sustainable revenue stream. Using hackintosh version of MacOS 11 to current(MacOS 26), for those who dont know, Hackintosh is just a illeagal copy of Macintosh, or the Macbooks we use, so they need the MacOS that can run on sillicon other than Apple’s sillicon so thats why they are like cracked and all. hopefully I am not banned for distributing kindof illeagal information so here is a disclaimer: educational purppouses only here is a screenshot of me trying to debug why GravityAds are not being shown(console is empty and the adslots are not there meaning something is failing silently)"}, {"post_id": 25640, "ts": "2026-07-10T01:35:38Z", "hours": 3.0, "text": "I dont get it I understand the 30% AI rule but you have to see my project is not AI slop, Its quality work. like high quality work. I am here training my own AI(new skill), and creating an AI platform which will likely enter the AI race. and im getting rejected for “ai slop”??? please look into this, also, can someone see my ship, why does it have only 35hrs logges, it should be 200+ hours. request the community to look into this."}, {"post_id": 25287, "ts": "2026-07-09T14:00:50Z", "hours": 11.0, "text": "FINALLY GOT THE EXE CODE TO COMPILE i put it in h1 because it is news that deserves h1, been batteling it for like 4 days my ship which i have sent for recertification for the 5th or 6th time is still under review Please remove this hard limit of not logging more than 10 hours in 1 go, like I cant be posting devlogs every nhow and then the exe is made, but not launching, debugging that now and with this, I am officially over 200 hours. YAYYY here is the screenshot of teh github actions completing succesfully(the .exe code compiling) also my Work on AI is still ongoing, My previous run failed my expecatitions, thistime the overall adtasets used is a bit too large(100gb+) and the slitting database code is causing errors."}, {"post_id": 25078, "ts": "2026-07-09T01:30:41Z", "hours": 17.0, "text": "and while I was thinking to upgrade the UI, I broke the .exe, like the exe is now broken. yay also my AMD credits request got approves so now I need to find a way to put my AI on AMD and such that the AI only gets activated when someone requests so that $50 lasts a long time, but in this system, Every user would have to wait for like 3 to 6 minutes for the GPU to turn on and boot, causing massive delays. Here is a screenshoot of the Github Actions failing."}, {"post_id": 24049, "ts": "2026-07-07T10:45:39Z", "hours": 29.0, "text": "Finally got th exe to open, though it has more bugs than the entire insect population, but atleast it works. thats all I care abt, here is the login Popup that is done. gotta make the UI better and fix all these bugs now. and this is only windows, I still have to do this for macOS and linux(if possible) and then publish the app on microsoft app store and apple app store"}, {"post_id": 22188, "ts": "2026-07-04T14:53:01Z", "hours": 6.0, "text": "put AI declaration, and got scammed 3 times by different websites, claiming they give free GPU hours but ended up giving nothing but takng my email. currently working on a way to host my model, working on the exact workflow that my next AI project under thalamus will use, it will be JARVIS so here is a spoiler. it will beat fable 5. and I have ran out of the weekly Claude limit on all my 5 email IDs. the instaler is still ended up being corrupt or incorrect. neither me nor AI can understand the problem here. and today i drank 3 cups of coffee instead f my usual 2. here is a screenshot of the error im recieving. and now i will send my shit for recertification again, now I have everything in place, including the AI declarations. hopefully, i wont hve to do it for the fifth or sixth time."}, {"post_id": 21434, "ts": "2026-07-03T11:53:33Z", "hours": 21.0, "text": "i will be surprised if i Dont get a migrane tommorow morning cause ive been spamming the .exe to work, the build for the .exe installer and .msi installer is failing again and again. to the shipwrights, the .exe is completely safe and virus free, there are chances that you will install a .msi instead of .exe on running, there are chances that the installer will fail, if it does, I know that it is failing and I am trying to fix it, if it works, then it works. heres a screenshot of the build failing. all I hope is that my ship gets verified this tme, 4rth time sending this bad boy for re cert."}, {"post_id": 19984, "ts": "2026-07-01T13:41:02Z", "hours": 9.0, "text": "for the Ship reviewing team sry for last re cert request, forgot that I have to reqrite m readme.md did that, now im requesting re cert again. hopeefuylly it passes this time. ps.my credits request is still pending, AMD relied that they will respond ASAP, it is under review(do companes do that) and then also sent an apology letter for the delay(why? why did they send it?). and please remove the compulsion of having a photo/ss for software projects."}, {"post_id": 19160, "ts": "2026-06-30T12:31:19Z", "hours": 8.0, "text": "I will now request Recertification, I think that I have done enough. To the reviewers, there are chances that you will get to use my self trained model if my AMD clloud credits request gets approved or my Google Cloud Instance quota gets increased, otherwise, the only change I have made is nothing. Im unable to put my AI online. SS uploaded is my kaggle account o you can see proof of AI model."}, {"post_id": 18445, "ts": "2026-06-29T16:35:31Z", "hours": 0.0, "text": "devlog: the 100-hour milestone, a humaneval script blowout, and parsing raw logs context recap (the architecture so far) for anyone catching up on the architecture overhaul: my initial ship for thalamus got rejected because the UI looked like an AI wrapper template. instead of shifting pixels, i decided to completely redo the backend engine. i’m training a custom, specialized 32B model optimized explicitly for programming logic—named thalamus-thor-1 . general fluff is scrubbed to save weight; a gemini researcher agent sub-team handles web scraping to cover info blind spots. resource-wise, it’s a tight squeeze. puter.js failed on orchestration, so i’m entirely dependent on a drying runway of AWS credits while waiting on google cloud and cheap amd gpu server approvals. with school starting in two days, i’ve also been preparing a lightweight remote framework to monitor logs and trigger test suites from school lab computers during short periods. earlier tonight, i managed to bypass kaggle’s aggressive session killers, successfully compiled the FP8/4-bit compressed weights on their tpu clusters, and immediately threw the model straight into openai’s standard 164-problem HumanEval benchmark. here is the live update from late tonight. milestone: 100 hours in the dirt first off, a quick personal timestamp: as of setting up these evaluation scripts, i have officially crossed the 100-hour mark of active development on this project. hitting triple digits while drowning in memory leaks, credit run-outs, and hardware timeouts is a brutal milestone, but the core engine is standing. the humaneval script blowout i watched the model successfully plow past problem 90, and it kept charging all the way through the end of the execution run. but right at the finish line, instead of outputting the final pass@1 score summary, the script threw a fatal calculation error and crashed out. the evaluation script generated the completions, ran the test assertions, and executed the functional verification, but the section of the code responsible for processing the final combination formula or compiling the macro-averaged percentage broke completely. standard human-eval metric scripts use an unbiased estimator calculation to evaluate pass@k: $$\\text{pass@k} = 1 - \\frac{\\binom{n-c}{k}}{\\binom{n}{k}}$$ if the array containing the total number of generated completions ($n$) or the correct passes ($c$) contains a single malformed row, a zero-division error, or an unhandled null exception from an execution timeout, the entire summary wrapper explodes. i’m staring at a broken metrics calculator, but the actual data isn’t lost. shifting to manual validation i’m not letting a broken parsing script mask the actual performance metrics of the weights after 100 hours of grind. the script managed to compile and save the fine-grained data out to the raw <input_path>_results.jsonl log file before the summary function choked. each row in that file contains the literal execution results marked cleanly as \"passed\" , \"failed\" , or \"timed out\" . instead of trying to debug the evaluator framework’s entry points tonight, i’m going to pull the raw JSON lines into a basic python script i’m writing right now. i’ll manually iterate through the rows, count the hard instances of \"passed\" across the tasks, and divide it by 164 to find the true, absolute pass@1 score myself. the raw logs are sitting in the directory. writing the cleanup parser now to see what thalamus-thor-1 actually scored."}, {"post_id": 18433, "ts": "2026-06-29T16:17:02Z", "hours": 1.0, "text": "devlog: impatient engineering, skipping the wait, and running humaneval live the project recap quick context if you’re catching up: my initial submission for thalamus got rejected because the UI looked like a boilerplate wrapper. instead of a superficial skin-deep fix, i decided to overhaul the entire backend engine and train a custom, hyper-specialized 32B model optimized strictly for heavy coding logic—named thalamus-thor-1 . conversational filler was stripped entirely from the datasets to keep the weights lean; my existing gemini agent team handles live web scraping to cover general knowledge blind spots. the resource situation is a major scramble. puter.js was too limited for my 9-agent orchestration, so i abandoned it entirely. i’m burning through the absolute last of my AWS credits while waiting on approvals for google cloud and cheap amd gpu clusters. plus, school starts in a couple days, so i’m building a lightweight remote execution layer to trigger test suites and monitor terminal logs from the school computer lab during our free periods. earlier today, after a brutal 13-hour marathon fighting out-of-memory (OOM) errors and kaggle’s session killers on deepspeed ZeRO-3 initialization spikes, i finally pushed the parameters under the threshold and successfully compiled the weights on their tpu v5e-8 kernels. here is the live update from 1.5 hours later. the midnight deployment: saying screw it the plan from my last post was to sleep on the weights and stand up the evaluation harness tomorrow morning. i lasted twenty minutes. when you spend 13 straight hours fighting initialization crashes just to get a model cooked, you don’t leave the weights sitting cold on a drive. the curiosity and momentum make it impossible to step away. i went ahead, pulled down the evaluation scripts, configured the isolation wrappers, and slammed the model directly into the testing execution loop. right now, thalamus-thor-1 is running a live trial by fire. it is chewing through openai’s standard HumanEval dataset in real-time. as of typing this, the model has already plowed past 90 out of the 164 total programming problems in the suite. the screenshot attached is the raw stdout window of my evaluation terminal. you can see the script compiling generations, feeding function signatures, and executing validation unit tests back-to-back. what the logs are proving seeing the terminal scroll through 90+ consecutive problems without dropping a syntax exception or throwing infinite token loops is an immense relief. when you compress a 32B model down to tight FP8 and 4-bit configurations while scaling the context matrix toward an 800K target using RoPE and YaRN embeddings, the mathematical margin for error is razor-thin. one wrong scale factor and the model outputs broken brackets or repeats strings indefinitely. instead, the model is executing functional correctness cleanly. it’s handling complex array manipulations, algorithmic logic, and nested structural filters completely raw. because the weights are completely unburdened by chat filler, it behaves less like a chatbot and more like a lightning-fast logic compiler—reading the docstring and immediately piping clean python code directly to the testing harness. once it clears problem 164, the system will output the final pass@1 percentage. that single number will tell the definitive truth about the token economics i calculated earlier. if the pass rate holds up against these unit tests, we are looking at a model that can genuinely compete with sonnet 4.6 on pure engineering tasks, while operating at an estimated internal compute cost of just $2 per million input tokens and $6 per million output tokens—giving my agent pipeline high-tier logic execution at the price of haiku 4.5. i’m staying up until the script hits 164. final data drops the second the testing loops finish executing."}, {"post_id": 18372, "ts": "2026-06-29T14:36:24Z", "hours": 16.0, "text": "devlog: thalamus-thor-1 is trained. recap, kaggle breakthrough, and humaneval context recap if you haven’t been tracking: my first ship attempt for thalamus got rejected because the UI looked like a generic boilerplate wrapper. instead of a superficial patch, i decided to completely overhaul the core engine and train a custom 32B model optimized strictly for raw code logic—named thalamus-thor-1 . general knowledge was stripped out during data prep to keep weights lean; my existing gemini-led researcher team handles live web scraping instead. infrastructure-wise, it’s a massive scramble. puter.js proved way too limited for my 9-agent pipeline orchestration, so i’m entirely stranded there, burning through the absolute last of my aws credits and waiting on gcp/amd approvals. also, school starts in two days, so i’m building a lightweight remote pipeline to trigger test suites and monitor logs straight from school lab machines during free periods. the kaggle breakthrough yesterday, i was stuck in deepspeed ZeRO-3 initialization hell. every time the notebook tried to load the base parameters on kaggle’s free tpu v5e-8 clusters, a massive compute spike would trigger their aggressive backend session killers, immediately dropping the kernel. i spent hours adjusting chunk prefills, gradient accumulation steps, and stripping background overhead until the initialization sequence squeezed just under the termination threshold. the run finally executed, completed its QLoRA fine-tuning passes at FP8, scaled the context toward an 800K target using RoPE and YaRN with 4-bit compression, and successfully compiled the weights. the screenshot attached to this update is the actual kaggle notebook page of my AI training notebook showing the setup. initial sanity checks compiling weights is only half the battle. when you aggressively compress a 32B model down to FP8/4-bit formats while applying long-context scaling embeddings, the positional math can warp. if it breaks, the model turns into a hallucination engine. i spun up a local instance to run direct logic checks, and the structural syntax held together perfectly. it didn’t choke on deeply nested loops, accurately referenced parameters defined deep inside large codebase contexts, and skipped conversational fluff entirely to jump straight into code execution. tomorrow’s trial: the humaneval protocol vibe checks aren’t enough. tomorrow morning, i am setting up an automated evaluation harness to run thalamus-thor-1 through the full OpenAI HumanEval dataset (164 coding problems) to calculate the raw pass@1 rate. this will prove the token economics math i did earlier: whether this custom setup can genuinely compete with sonnet 4.6 and opus 4.6 on pure engineering tasks while running at an estimated cost of $2/1M input and $6/1M output tokens (top-tier intelligence at the price of haiku 4.5). hosting is still the looming issue. permanently serving this with an extended context window requires a dedicated 80gb VRAM prompt server configuration, which i’m trying to source while my teammate builds out the social media presence to push links. on a side note, fighting package conflicts during this nightmare inspired a lightweight side-utility: a clean CLI tool for coders that natively resolves dependency hell. the core engine is working perfectly, and you can track it here: https://stardance.hackclub.com/projects/6323 evaluation metrics drop tomorrow morning. let’s see what the numbers show."}, {"post_id": 17483, "ts": "2026-06-28T08:58:36Z", "hours": 13.0, "text": "quick recap on why the scope completely escalated: my first ship attempt for thalamus got rejected because the UI looked like a generic template wrapper and the documentation felt AI-generated. instead of just patching the CSS, i decided to completely overhaul the core engine and train a custom 32B model optimized for raw code logic—named thalamus-thor-1 . here is the brutal reality of the last 24 hours: 1. the infrastructure and credit crunch i previously planned to use puter.js as a temporary backend fallback router for the claude models while training. scratch that. puter.js is way too limited for the multi-agent architecture i’m running, so i’m entirely stranded there. right now, i’m burning through the absolute last of my AWS credits. i’ve applied for google cloud credits to spin up proper GPU instances, but everything hinges on getting that approval before the AWS runway hits zero. surviving on nothing but hope and coffee. 2. 13 hours battling kaggle i spent 13 hours straight fighting out-of-memory (OOM) errors and kaggle’s infrastructure quirks. i successfully patched the local OOM issues, but now i’m hitting a wall with their session killers. the platform completely terminates the notebook due to compute spikes right when ZeRO-3 is about to initialize. the screenshot for this post is the raw kaggle output cell showing exactly where the infra drops the hammer. because of this, i’ve applied for amd gpu credits. they’re vastly cheaper than nvidia options but offer massive amounts of vram, which is exactly what a high-context 32b setup needs to handle codebase ingestions via RoPE and YaRN. 3. Social Media on the startup side, my teammate is taking over the social media presence and will start pushing updates the second the app is not hope it’s functional. 4. dependency utility side-project ironically, fighting package conflicts during this 13-hour nightmare gave me an idea for a quality-of-life tool. i built a clean CLI tool for coders that natively resolves dependency hell. the core logic is already up and running perfectly. you can track and follow the new CLI project here: https://stardance.hackclub.com/projects/6323 going back to tweaking the zero-3 configurations before the credits completely empty out."}, {"post_id": 16791, "ts": "2026-06-27T08:04:57Z", "hours": 0.0, "text": "Just saw that puter.js brand themselves as unlimited AI, they arrent actually unlimited. Now on my hunt to find something to temporarily replace Puter.js for claude requests. Gotta start training my AI fast."}, {"post_id": 16771, "ts": "2026-06-27T07:22:32Z", "hours": 5.0, "text": "finally back from my trip. i didn’t post the last update because the entire scope of what i’m doing just completely escalated. to recap: my first ship attempt for thalamus got rejected. the feedback was that the UI looked too much like a generic AI wrapper boilerplate, the README felt AI-generated, and it needed unique features. instead of just doing a superficial skin change or tweaking the CSS to look pretty, i am using this as a reason to completely redo the underlying engineering and turn this into a true L4.5 or L5 agent that destroys things like claude code. i attached a photo of my handwritten notes to this post. it maps out the exact problem statements i’m tackling and the major flaws of currently existing solutions out there. my handwriting is messy, but the technical breakdown is all there. here is the actual game plan for the AI routing and training infrastructure: right now, i am updating the app to route claude models via puter.js. to be clear, this isn’t for the user interface—puter.js is strictly my temporary backend AI router. it acts as a fallback bridge while i build out my own custom models. i’m going to benchmark my self-trained models against every single task in my 9-agent pipeline. until my model is 100% foolproof on a specific task, puter.js will keep routing that specific agent’s requests to claude. but the second my custom model qualifies and beats the benchmarks for code mode, puter.js stops getting requests for it and the pipeline switches over entirely to my model. the actual training project is running parallel to this: i’m fine-tuning a 32B model optimized explicitly for pure coding and logical tasks. training is happening via QLoRA at FP8 with a 4k/8k base context, but i’m extending the context window all the way up to 800K using RoPE and YaRN with 4-bit compression, flashattention-3, and chunked prefills so it can digest entire codebases at once. i’m doing the actual training on kaggle kernels using the 30 free weekly hours on their TPU v5e-8 clusters. based on my calculations, this model will completely fail at things like history or literature, but it doesn’t need to be good at them. my gemini researcher agent team already handles live web search and pulls fresh information for those blind spots anyway. on pure code, this thing should easily beat sonnet 4.6 and compete with opus 4.6. the token economics are insane if this works: about $2 per million input tokens and $6 per million output tokens based on max output possible per hour versus GPU compute costs. that gives me sonnet/opus level intelligence at the price of haiku 4.5. the real bottleneck right now isn’t the data or the training—it’s hosting. deploying a 32B model with an extended context window means i need to secure an 80gb VRAM per parellel prompt configuration. figuring out how to source or fund that infrastructure without incinerating cash is the main issue i’m working through while compiling the training scripts. Today i will just do the rerouting to puter.js, then start working on fine tuning the 32b model. Wish me luck."}, {"post_id": 12491, "ts": "2026-06-20T12:13:17Z", "hours": 8.0, "text": "Following the rejection of my initial shipping submission, I am pivoting to address the required refinements. I am viewing this as a system audit rather than a failure—the core engineering is stable, but the interface and documentation need to move closer to a production-grade tool. Engineering Status Module Current Status Notes 9-Agent Pipeline Stable Successfully debugged AgentRouter integration. VM Bridge Alpha QEMU bridge is operational; optimizing latency. Claude Code Workflow Ready Pipeline errors resolved; full parsing functional. Feedback Audit The review team highlighted specific areas for improvement: Interface: The current UI requires a shift from “AI-template” style to a proprietary “Control Center” aesthetic. Documentation: The README is currently AI-generated; I will rewrite it to reflect the technical architecture of the 9-agent pipeline and the VM bridge implementation. Utility: Implementing “startup-specific” features that move beyond a standard chat wrapper. Schedule & Roadmap I am entering a maintenance window due to travel. No commits will be pushed during this period. June 20: Finalizing technical walkthrough assets. June 21 - 25: OFFLINE (Maintenance Mode) . June 26: Resuming development. Priority: UI Refactor & Documentation Audit. Project Repository: https://github.com/hardcoregamingsyle/thalamus ps. for anyone wondering, use markdown to format your devposts"}, {"post_id": 11911, "ts": "2026-06-19T12:32:42Z", "hours": 10.0, "text": "today has to be my most unproductive day yet. I am vey frustrated because of some error, and when I give it to claude code, guess what, Claude code is getting errors, Like it is topping in between, or giving me error 400, 401 or 403, That is what Im doing right now. it is just not working. Like I was fixing the Agent Output parasing so the Commands are Read properly. Thought to refractor all big files, Now refractoring myself would be time consuming, so I went to claude code, and giess what, Claude code is giving me errors, And i ended up spending the same number of time working as if I wouldve done manually. but atleast my code is fine and does not have any bugs. Once the WebUI is fixed PC, I will work on WebUI for other Devices, then work on the .exe file for the VM for any OS, I havegathered the ISOs for Windows 10, Widnows 11(Both are Pro Vesions and Preactivated), MacOS 11to MacOS 16, Android 14 and Android 13, Linux, Ubuntu and other Open Source OS systems, aftre that I will work on iOS. which is byfar the most difficult one. I also have to make some changes to the Referal System and Token Deduction system, It is currently Highly overpriced. Will update yall once I make some more progress"}, {"post_id": 11346, "ts": "2026-06-18T15:39:37Z", "hours": 6.0, "text": "This is devlog so I dont think we should only send our app progressions but treat it like a communuty, so to start with it. I finally got rid of a 400 error, found out while getting rid of it, my codebase got infected with corrupted files, I am now using claude code to clean uo my codebase. also sadly, I had to delete my entir git history, which is very sad. I am now from 68 coomits to just 4(the first commit has +~70k lines as proof that i had real work). Anyways, heres a photo of 2 corrupted files which I manually found. I better get some sleep before I get a headache and weaken my eyes."}, {"post_id": 11182, "ts": "2026-06-18T09:05:52Z", "hours": 8.0, "text": "Today, I finally got the code mode and study mode to work, They are a bit slow, Im working on it, also. right now, Im busy debugging this error that Im getting by claude code, all help will be highly appriciated. My Project is ready to be shipped, Will ship it today, all that I need to do is improve the already good documentation. REQUEST: Please help me solve this error that I am again and again getting while using claude code, I changed the settings.json of the .claude folder so I can use Claude code but the API will be of Agentrouter.org so I dont hve to buy the claude Plans. It works usually but sometimes I get this error, and when I di get it, it is just annoying. Demo: https://thalamus.aphantic.skinticals.com You will have to log in for More usage and access to Code mode and Research mode, had to restrict because it costs me money whenever someone uses it."}, {"post_id": 10641, "ts": "2026-06-17T14:31:46Z", "hours": 26.0, "text": "Well because my origional vibe coder got aquired by freebuff, I spent the rest of this time migrating from vly to Claude Code, made some progress, In initial testing, I found out that code mode and study mode were broken, Working on that, also stepping up my .exe from a Electron wrapper to a natice C++, C# Desktop app, Like of Minecraft/Roblox. This is taking me some time as claude code is kinda dumb for me because im like using it as my main Agent for the first time so it will take me some time getting the hang of it, Also I have cnnected Claude code to agentrouter, which is a third party that gives $100 initial credits and then $25 a day for signing in. Using my fav multi account system to get the most of this also. Will post another devlog as soon as I make some more progress. The image uploaded is just me working on code mode."}, {"post_id": 2023, "ts": "2026-06-04T09:34:49Z", "hours": 0.0, "text": "fixing the sandbox to actually work"}]...
|
Title: Thalamus + AgentOverflow by @Hardcore_Real_ Title: Thalamus + AgentOverflow by @Hardcore_Real_Name_is_nitish | Stardance
Devlogs:
fixing the sandbox to actually work
Well because my origional vibe coder got aquired by freebuff, I spent the rest of this time migrating from vly to Claude Code, made some progress, In initial testing, I found out that code mode and study mode were broken, Working on that, also stepping up my .exe from a Electron wrapper to a natice C++, C# Desktop app, Like of Minecraft/Roblox. This is taking me some time as claude code is kinda dumb for me because im like using it as my main Agent for the first time so it will take me some time getting the hang of it, Also I have cnnected Claude code to agentrouter, which is a third party that gives $100 initial credits and then $25 a day for signing in. Using my fav multi account system to get the most of this also. Will post another devlog as soon as I make some more progress. The image uploaded is just me working on code mode.
Today, I finally got the code mode and study mode to work, They are a bit slow, Im working on it, also. right now, Im busy debugging this error that Im getting by claude code, all help will be highly appriciated. My Project is ready to be shipped, Will ship it today, all that I need to do is improve the already good documentation. REQUEST: Please help me solve this error that I am again and again getting while using claude code, I changed the settings.json of the .claude folder so I can use Claude code but the API will be of Agentrouter.org so I dont hve to buy the claude Plans. It works usually but sometimes I get this error, and when I di get it, it is just annoying. Demo: https://thalamus.aphantic.skinticals.com You will have to log in for More usage and access to Code mode and Research mode, had to restrict because it costs me money whenever someone uses it.
This is devlog so I dont think we should only send our app progressions but treat it like a communuty, so to start with it. I finally got rid of a 400 error, found out while getting rid of it, my codebase got infected with corrupted files, I am now using claude code to clean uo my codebase. also sadly, I had to delete my entir git history, which is very sad. I am now from 68 coomits to just 4(the first commit has +~70k lines as proof that i had real work). Anyways, heres a photo of 2 corrupted files which I manually found. I better get some sleep before I get a headache and weaken my eyes.
today has to be my most unproductive day yet. I am vey frustrated because of some error, and when I give it to claude code, guess what, Claude code is getting errors, Like it is topping in between, or giving me error 400, 401 or 403, That is what Im doing right now. it is just not working. Like I was fixing the Agent Output parasing so the Commands are Read properly. Thought to refractor all big files, Now refractoring myself would be time consuming, so I went to claude code, and giess what, Claude code is giving me errors, And i ended up spending the same number of time working as if I wouldve done manually. but atleast my code is fine and does not have any bugs. Once the WebUI is fixed PC, I will work on WebUI for other Devices, then work on the .exe file for the VM for any OS, I havegathered the ISOs for Windows 10, Widnows 11(Both are Pro Vesions and Preactivated), MacOS 11to MacOS 16, Android 14 and Android 13, Linux, Ubuntu and other Open Source OS systems, aftre that I will work on iOS. which is byfar the most difficult one. I also have to make some changes to the Referal System and Token Deduction system, It is currently Highly overpriced. Will update yall once I make some more progress
Following the rejection of my initial shipping submission, I am pivoting to address the required refinements. I am viewing this as a system audit rather than a failure—the core engineering is stable, but the interface and documentation need to move closer to a production-grade tool. Engineering Status Module Current Status Notes 9-Agent Pipeline Stable Successfully debugged AgentRouter integration. VM Bridge Alpha QEMU bridge is operational; optimizing latency. Claude Code Workflow Ready Pipeline errors resolved; full parsing functional. Feedback Audit The review team highlighted specific areas for improvement: Interface: The current UI requires a shift from “AI-template” style to a proprietary “Control Center” aesthetic. Documentation: The README is currently AI-generated; I will rewrite it to reflect the technical architecture of the 9-agent pipeline and the VM bridge implementation. Utility: Implementing “startup-specific” features that move beyond a standard chat wrapper. Schedule & Roadmap I am entering a maintenance window due to travel. No commits will be pushed during this period. June 20: Finalizing technical walkthrough assets. June 21 - 25: OFFLINE (Maintenance Mode) . June 26: Resuming development. Priority: UI Refactor & Documentation Audit. Project Repository: https://github.com/hardcoregamingsyle/thalamus ps. for anyone wondering, use markdown to format your devposts
finally back from my trip. i didn’t post the last update because the entire scope of what i’m doing just completely escalated. to recap: my first ship attempt for thalamus got rejected. the feedback was that the UI looked too much like a generic AI wrapper boilerplate, the README felt AI-generated, and it needed unique features. instead of just doing a superficial skin change or tweaking the CSS to look pretty, i am using this as a reason to completely redo the underlying engineering and turn this into a true L4.5 or L5 agent that destroys things like claude code. i attached a photo of my handwritten notes to this post. it maps out the exact problem statements i’m tackling and the major flaws of currently existing solutions out there. my handwriting is messy, but the technical breakdown is all there. here is the actual game plan for the AI routing and training infrastructure: right now, i am updating the app to route claude models via puter.js. to be clear, this isn’t for the user interface—puter.js is strictly my temporary backend AI router. it acts as a fallback bridge while i build out my own custom models. i’m going to benchmark my self-trained models against every single task in my 9-agent pipeline. until my model is 100% foolproof on a specific task, puter.js will keep routing that specific agent’s requests to claude. but the second my custom model qualifies and beats the benchmarks for code mode, puter.js stops getting requests for it and the pipeline switches over entirely to my model. the actual training project is running parallel to this: i’m fine-tuning a 32B model optimized explicitly for pure coding and logical tasks. training is happening via QLoRA at FP8 with a 4k/8k base context, but i’m extending the context window all the way up to 800K using RoPE and YaRN with 4-bit compression, flashattention-3, and chunked prefills so it can digest entire codebases at once. i’m doing the actual training on kaggle kernels using the 30 free weekly hours on their TPU v5e-8 clusters. based on my calculations, this model will completely fail at things like history or literature, but it doesn’t need to be good at them. my gemini researcher agent team already handles live web search and pulls fresh information for those blind spots anyway. on pure code, this thing should easily beat sonnet 4.6 and compete with opus 4.6. the token economics are insane if this works: about $2 per million input tokens and $6 per million output tokens based on max output possible per hour versus GPU compute costs. that gives me sonnet/opus level intelligence at the price of haiku 4.5. the real bottleneck right now isn’t the data or the training—it’s hosting. deploying a 32B model with an extended context window means i need to secure an 80gb VRAM per parellel prompt configuration. figuring out how to source or fund that infrastructure without incinerating cash is the main issue i’m working through while compiling the training scripts. Today i will just do the rerouting to puter.js, then start working on fine tuning the 32b model. Wish me luck.
Just saw that puter.js brand themselves as unlimited AI, they arrent actually unlimited. Now on my hunt to find something to temporarily replace Puter.js for claude requests. Gotta start training my AI fast.
quick recap on why the scope completely escalated: my first ship attempt for thalamus got rejected because the UI looked like a generic template wrapper and the documentation felt AI-generated. instead of just patching the CSS, i decided to completely overhaul the core engine and train a custom 32B model optimized for raw code logic—named thalamus-thor-1 . here is the brutal reality of the last 24 hours: 1. the infrastructure and credit crunch i previously planned to use puter.js as a temporary backend fallback router for the claude models while training. scratch that. puter.js is way too limited for the multi-agent architecture i’m running, so i’m entirely stranded there. right now, i’m burning through the absolute last of my AWS credits. i’ve applied for google cloud credits to spin up proper GPU instances, but everything hinges on getting that approval before the AWS runway hits zero. surviving on nothing but hope and coffee. 2. 13 hours battling kaggle i spent 13 hours straight fighting out-of-memory (OOM) errors and kaggle’s infrastructure quirks. i successfully patched the local OOM issues, but now i’m hitting a wall with their session killers. the platform completely terminates the notebook due to compute spikes right when ZeRO-3 is about to initialize. the screenshot for this post is the raw kaggle output cell showing exactly where the infra drops the hammer. because of this, i’ve applied for amd gpu credits. they’re vastly cheaper than nvidia options but offer massive amounts of vram, which is exactly what a high-context 32b setup needs to handle codebase ingestions via RoPE and YaRN. 3. Social Media on the startup side, my teammate is taking over the social media presence and will start pushing updates the second the app is not hope it’s functional. 4. dependency utility side-project ironically, fighting package conflicts during this 13-hour nightmare gave me an idea for a quality-of-life tool. i built a clean CLI tool for coders that natively resolves dependency hell. the core logic is already up and running perfectly. you can track and follow the new CLI project here: https://stardance.hackclub.com/projects/6323 going back to tweaking the zero-3 configurations before the credits completely empty out.
devlog: thalamus-thor-1 is trained. recap, kaggle breakthrough, and humaneval context recap if you haven’t been tracking: my first ship attempt for thalamus got rejected because the UI looked like a generic boilerplate wrapper. instead of a superficial patch, i decided to completely overhaul the core engine and train a custom 32B model optimized strictly for raw code logic—named thalamus-thor-1 . general knowledge was stripped out during data prep to keep weights lean; my existing gemini-led researcher team handles live web scraping instead. infrastructure-wise, it’s a massive scramble. puter.js proved way too limited for my 9-agent pipeline orchestration, so i’m entirely stranded there, burning through the absolute last of my aws credits and waiting on gcp/amd approvals. also, school starts in two days, so i’m building a lightweight remote pipeline to trigger test suites and monitor logs straight from school lab machines during free periods. the kaggle breakthrough yesterday, i was stuck in deepspeed ZeRO-3 initialization hell. every time the notebook tried to load the base parameters on kaggle’s free tpu v5e-8 clusters, a massive compute spike would trigger their aggressive backend session killers, immediately dropping the kernel. i spent hours adjusting chunk prefills, gradient accumulation steps, and stripping background overhead until the initialization sequence squeezed just under the termination threshold. the run finally executed, completed its QLoRA fine-tuning passes at FP8, scaled the context toward an 800K target using RoPE and YaRN with 4-bit compression, and successfully compiled the weights. the screenshot attached to this update is the actual kaggle notebook page of my AI training notebook showing the setup. initial sanity checks compiling weights is only half the battle. when you aggressively compress a 32B model down to FP8/4-bit formats while applying long-context scaling embeddings, the positional math can warp. if it breaks, the model turns into a hallucination engine. i spun up a local instance to run direct logic checks, and the structural syntax held together perfectly. it didn’t choke on deeply nested loops, accurately referenced parameters defined deep inside large codebase contexts, and skipped conversational fluff entirely to jump straight into code execution. tomorrow’s trial: the humaneval protocol vibe checks aren’t enough. tomorrow morning, i am setting up an automated evaluation harness to run thalamus-thor-1 through the full OpenAI HumanEval dataset (164 coding problems) to calculate the raw pass@1 rate. this will prove the token economics math i did earlier: whether this custom setup can genuinely compete with sonnet 4.6 and opus 4.6 on pure engineering tasks while running at an estimated cost of $2/1M input and $6/1M output tokens (top-tier intelligence at the price of haiku 4.5). hosting is still the looming issue. permanently serving this with an extended context window requires a dedicated 80gb VRAM prompt server configuration, which i’m trying to source while my teammate builds out the social media presence to push links. on a side note, fighting package conflicts during this nightmare inspired a lightweight side-utility: a clean CLI tool for coders that natively resolves dependency hell. the core engine is working perfectly, and you can track it here: https://stardance.hackclub.com/projects/6323 evaluation metrics drop tomorrow morning. let’s see what the numbers show.
devlog: impatient engineering, skipping the wait, and running humaneval live the project recap quick context if you’re catching up: my initial submission for thalamus got rejected because the UI looked like a boilerplate wrapper. instead of a superficial skin-deep fix, i decided to overhaul the entire backend engine and train a custom, hyper-specialized 32B model optimized strictly for heavy coding logic—named thalamus-thor-1 . conversational filler was stripped entirely from the datasets to keep the weights lean; my existing gemini agent team handles live web scraping to cover general knowledge blind spots. the resource situation is a major scramble. puter.js was too limited for my 9-agent orchestration, so i abandoned it entirely. i’m burning through the absolute last of my AWS credits while waiting on approvals for google cloud and cheap amd gpu clusters. plus, school starts in a couple days, so i’m building a lightweight remote execution layer to trigger test suites and monitor terminal logs from the school computer lab during our free periods. earlier today, after a brutal 13-hour marathon fighting out-of-memory (OOM) errors and kaggle’s session killers on deepspeed ZeRO-3 initialization spikes, i finally pushed the parameters under the threshold and successfully compiled the weights on their tpu v5e-8 kernels. here is the live update from 1.5 hours later. the midnight deployment: saying screw it the plan from my last post was to sleep on the weights and stand up the evaluation harness tomorrow morning. i lasted twenty minutes. when you spend 13 straight hours fighting initialization crashes just to get a model cooked, you don’t leave the weights sitting cold on a drive. the curiosity and momentum make it impossible to step away. i went ahead, pulled down the evaluation scripts, configured the isolation wrappers, and slammed the model directly into the testing execution loop. right now, thalamus-thor-1 is running a live trial by fire. it is chewing through openai’s standard HumanEval dataset in real-time. as of typing this, the model has already plowed past 90 out of the 164 total programming problems in the suite. the screenshot attached is the raw stdout window of my evaluation terminal. you can see the script compiling generations, feeding function signatures, and executing validation unit tests back-to-back. what the logs are proving seeing the terminal scroll through 90+ consecutive problems without dropping a syntax exception or throwing infinite token loops is an immense relief. when you compress a 32B model down to tight FP8 and 4-bit configurations while scaling the context matrix toward an 800K target using RoPE and YaRN embeddings, the mathematical margin for error is razor-thin. one wrong scale factor and the model outputs broken brackets or repeats strings indefinitely. instead, the model is executing functional correctness cleanly. it’s handling complex array manipulations, algorithmic logic, and nested structural filters completely raw. because the weights are completely unburdened by chat filler, it behaves less like a chatbot and more like a lightning-fast logic compiler—reading the docstring and immediately piping clean python code directly to the testing harness. once it clears problem 164, the system will output the final pass@1 percentage. that single number will tell the definitive truth about the token economics i calculated earlier. if the pass rate holds up against these unit tests, we are looking at a model that can genuinely compete with sonnet 4.6 on pure engineering tasks, while operating at an estimated internal compute cost of just $2 per million input tokens and $6 per million output tokens—giving my agent pipeline high-tier logic execution at the price of haiku 4.5. i’m staying up until the script hits 164. final data drops the second the testing loops finish executing.
devlog: the 100-hour milestone, a humaneval script blowout, and parsing raw logs context recap (the architecture so far) for anyone catching up on the architecture overhaul: my initial ship for thalamus got rejected because the UI looked like an AI wrapper template. instead of shifting pixels, i decided to completely redo the backend engine. i’m training a custom, specialized 32B model optimized explicitly for programming logic—named thalamus-thor-1 . general fluff is scrubbed to save weight; a gemini researcher agent sub-team handles web scraping to cover info blind spots. resource-wise, it’s a tight squeeze. puter.js failed on orchestration, so i’m entirely dependent on a drying runway of AWS credits while waiting on google cloud and cheap amd gpu server approvals. with school starting in two days, i’ve also been preparing a lightweight remote framework to monitor logs and trigger test suites from school lab computers during short periods. earlier tonight, i managed to bypass kaggle’s aggressive session killers, successfully compiled the FP8/4-bit compressed weights on their tpu clusters, and immediately threw the model straight into openai’s standard 164-problem HumanEval benchmark. here is the live update from late tonight. milestone: 100 hours in the dirt first off, a quick personal timestamp: as of setting up these evaluation scripts, i have officially crossed the 100-hour mark of active development on this project. hitting triple digits while drowning in memory leaks, credit run-outs, and hardware timeouts is a brutal milestone, but the core engine is standing. the humaneval script blowout i watched the model successfully plow past problem 90, and it kept charging all the way through the end of the execution run. but right at the finish line, instead of outputting the final pass@1 score summary, the script threw a fatal calculation error and crashed out. the evaluation script generated the completions, ran the test assertions, and executed the functional verification, but the section of the code responsible for processing the final combination formula or compiling the macro-averaged percentage broke completely. standard human-eval metric scripts use an unbiased estimator calculation to evaluate pass@k: $$\text{pass@k} = 1 - \frac{\binom{n-c}{k}}{\binom{n}{k}}$$ if the array containing the total number of generated completions ($n$) or the correct passes ($c$) contains a single malformed row, a zero-division error, or an unhandled null exception from an execution timeout, the entire summary wrapper explodes. i’m staring at a broken metrics calculator, but the actual data isn’t lost. shifting to manual validation i’m not letting a broken parsing script mask the actual performance metrics of the weights after 100 hours of grind. the script managed to compile and save the fine-grained data out to the raw <input_path>_results.jsonl log file before the summary function choked. each row in that file contains the literal execution results marked cleanly as "passed" , "failed" , or "timed out" . instead of trying to debug the evaluator framework’s entry points tonight, i’m going to pull the raw JSON lines into a basic python script i’m writing right now. i’ll manually iterate through the rows, count the hard instances of "passed" across the tasks, and divide it by 164 to find the true, absolute pass@1 score myself. the raw logs are sitting in the directory. writing the cleanup parser now to see what thalamus-thor-1 actually scored.
I will now request Recertification, I think that I have done enough. To the reviewers, there are chances that you will get to use my self trained model if my AMD clloud credits request gets approved or my Google Cloud Instance quota gets increased, otherwise, the only change I have made is nothing. Im unable to put my AI online. SS uploaded is my kaggle account o you can see proof of AI model.
for the Ship reviewing team sry for last re cert request, forgot that I have to reqrite m readme.md did that, now im requesting re cert again. hopeefuylly it passes this time. ps.my credits request is still pending, AMD relied that they will respond ASAP, it is under review(do companes do that) and then also sent an apology letter for the delay(why? why did they send it?). and please remove the compulsion of having a photo/ss for software projects.
i will be surprised if i Dont get a migrane tommorow morning cause ive been spamming the .exe to work, the build for the .exe installer and .msi installer is failing again and again. to the shipwrights, the .exe is completely safe and virus free, there are chances that you will install a .msi instead of .exe on running, there are chances that the installer will fail, if it does, I know that it is failing and I am trying to fix it, if it works, then it works. heres a screenshot of the build failing. all I hope is that my ship gets verified this tme, 4rth time sending this bad boy for re cert.
put AI declaration, and got scammed 3 times by different websites, claiming they give free GPU hours but ended up giving nothing but takng my email. currently working on a way to host my model, working on the exact workflow that my next AI project under thalamus will use, it will be JARVIS so here is a spoiler. it will beat fable 5. and I have ran out of the weekly Claude limit on all my 5 email IDs. the instaler is still ended up being corrupt or incorrect. neither me nor AI can understand the problem here. and today i drank 3 cups of coffee instead f my usual 2. here is a screenshot of the error im recieving. and now i will send my shit for recertification again, now I have everything in place, including the AI declarations. hopefully, i wont hve to do it for the fifth or sixth time.
Finally got th exe to open, though it has more bugs than the entire insect population, but atleast it works. thats all I care abt, here is the login Popup that is done. gotta make the UI better and fix all these bugs now. and this is only windows, I still have to do this for macOS and linux(if possible) and then publish the app on microsoft app store and apple app store
and while I was thinking to upgrade the UI, I broke the .exe, like the exe is now broken. yay also my AMD credits request got approves so now I need to find a way to put my AI on AMD and such that the AI only gets activated when someone requests so that $50 lasts a long time, but in this system, Every user would have to wait for like 3 to 6 minutes for the GPU to turn on and boot, causing massive delays. Here is a screenshoot of the Github Actions failing.
FINALLY GOT THE EXE CODE TO COMPILE i put it in h1 because it is news that deserves h1, been batteling it for like 4 days my ship which i have sent for recertification for the 5th or 6th time is still under review Please remove this hard limit of not logging more than 10 hours in 1 go, like I cant be posting devlogs every nhow and then the exe is made, but not launching, debugging that now and with this, I am officially over 200 hours. YAYYY here is the screenshot of teh github actions completing succesfully(the .exe code compiling) also my Work on AI is still ongoing, My previous run failed my expecatitions, thistime the overall adtasets used is a bit too large(100gb+) and the slitting database code is causing errors.
I dont get it I understand the 30% AI rule but you have to see my project is not AI slop, Its quality work. like high quality work. I am here training my own AI(new skill), and creating an AI platform which will likely enter the AI race. and im getting rejected for “ai slop”??? please look into this, also, can someone see my ship, why does it have only 35hrs logges, it should be 200+ hours. request the community to look into this.
Adding GravityAds for some revenue(this is a startup, remember), tried to add Payment but BuyMeACoffee crashed, HCB Bank isnt in india so I had to go with BMAC, so payments is the next task, the exe had some reworks done. still looking for a way to make my $50 of AMD credits last long, also I noticed that Rather than wasting it on running, I can use the aprox 200gb of VRAM to train a very powerful AI Model in like 25 hours, and then give you all the .gguf and then it is up to you if you want to test the AI or not. also, I am now applying for Emergent Ventures(EV) and Theil Fellowship, because I need the most amount of grant possible to get this to work. and currently on a hunt to find ISOs for the QEMU VM system, I am looking for Preactivated Pro verison of Windows 10 and 11, which I think is illigal but right now, we havent launched publically yet so I dont think it matters that I am going on the illeagal path of Software Piracy, but I promise I will once use real lisenced Windows version. For MacOS, it is something that can never be on the legal route till I get a good sustainable revenue stream. Using hackintosh version of MacOS 11 to current(MacOS 26), for those who dont know, Hackintosh is just a illeagal copy of Macintosh, or the Macbooks we use, so they need the MacOS that can run on sillicon other than Apple’s sillicon so thats why they are like cracked and all. hopefully I am not banned for distributing kindof illeagal information so here is a disclaimer: educational purppouses only here is a screenshot of me trying to debug why GravityAds are not being shown(console is empty and the adslots are not there meaning something is failing silently)
gravityads yet not working, BMAC under progress, how to deploy my AI is still a question with no practical answer, I asked my parents to buy huggingface paid plan for me so I can use zeroGPU, I got a hour long lecture on How I have to escape the fake tech world, and how these people scam us. currently working on improving the feel of the desktop app and vibe coder, step-1: Give it the feel of Claude Code, Codex or any other l4 vibe coder(easy) step 2: Upgrade the feel to make it a true l4.5 Agent(near impossible level difficulty, I wont be copying anyone here, I would have to make my own definatation of L4.5) Side Quest(Very Important): Either get yourself a Good amount of GPU hours or get yourself a way to deploy your AI model somewhere. Frustration is at it’s peak. and can someone please gift me a RTX Pro 6000(Blackwell), it has 96gb of VRAM, enough to trsin a 64B Model on it and make it very very good. here is the screenshot of a response from AI in the desktop App. Now I will work on MacOS desktop app.
I was trying to put in dynamic AI coding pipeline, and guess what, My code mode compeltely crashed. My AI is still being trained(errors o errors+ hitting my weekly limit.), also,My application for Emerging Ventures got rejected because I hadnt put in all info(mb), Will reapply again. For some reason, my GravityAds Account is showing unverified Via API but verified via dashboard, and I am still working on BMAC integration for paywall, I need to figure out on the payment and stuff. this is Devpost 1 of asking for the Claude Max x20 plan in the shop. and Huggingface GPU hours credits in the shop. here is a screenshot of of my rejection letter.
don’t ask whats going on with my self trained AI, it is in no good state, the code either works, or it doesnt, or it works but something is off. I have a new idea, under thalamus, I will have AgentOverflow AgentOverflow is like StackOverflow but for Agents, will explain it later(when I actually start to work on it) BMAC is in queue. Dynamic pipeline is kindof done, kindof undone. Desktop app turned out to be nicer than I thought. rest I have nothing to say. here is a screenshot of the desktop app build mode returning a valid output instead of an error thats all I have to say
working on Agentoberflow, which is stackoverlow for agents, Side by side working on dynamic pipeline. For agentoverflow, as a starting point, I am adding the entire stackoverflow database, like its entire codebase, Cleaned and filtered, Vectorised, and Graphed using GraphRAG, Storing it in maybe AWS EC2 as it is the cheapest. Also working on the Agentoverflow website, though both have the same database, like the exact same database in convex so that in the future, when I want to make a My entire ecosystem at 1 place, it is very very easy for me to do. and my life is in the greatest mess possible, my parents make me constantly change setups, due to which I need my own laptop, in the sense which I got on my own with my own effort so they cant take it from me, and I also need my own phone. so I dont have to make sure that my whatsapp is logged into every device. and I am completely dropping Payments in the App because BMAC requires stripe to connect payment. here is a screenshot of the new Agentoverflow Website.
more work on Agentoverflow, added MCP tool, currently around 10% of stackoveflow database is in this bad boy. here is a ss of what im thinking of doing next.
Testing MCP, I am setting AI training Aside because im currently creating the database. hopefully everythng will work, this AgentOverflow is like the brilliant idea, like barely any competition. website: https://agentoverflow.aphantic.skinticals.com/ git repo: https://github.com/hardcoregamingsyle/agentoverflow currently ingesting the entire stackoverflow database, while integrating it into thalamus. will also submit agentoverflow as a tool for ClaudeCode. SS of new website.
added 3d and scroll anims to Thalamus, increased the Limits per day in Agentoverflow so it can compete with StackOverflow, currently at 300k Examples for Agentoverflow, compute is slow, will reach to 3.7m+ by 3 weeks.
did the UI and animations for agentoverflow also. https://agentoverflow.aphantic.skinticals.com/
agentoverflow has 475k+ examples in dataset(total is 3.7m), and im so great, I forgot to git push and deleed my folder, great, time to redo the special feature(not releasing yet). I published my site to google analytics. Im talking with my school rn if they can take my project and make it the official school AI. my school’s beurocracy is slow as hell. so idk if it will ever happen. also, here is a ss of the admin panel(of agentoverflow)
Now working o SEO and indexing, Agentoverflow db has over 730k entries. Intigrated Agentoverflow MCP into thalamus and also integrated Sketchfab MCP for 3d models in case of Gameev or projects that require a 3d model. Currently testing my Code mode. Prompt: Build a highly advanced flight sim. Also, quick question, I made this AI, so I can use my own AI to build other projects, right.
i havent slept since the last 24 hours, cause FIFA world Cup was at night, and during the day, I never sleep, so Im forcing myself to work as hell. already crossedthe 800k point. Improved the Dispatcher in Thalamus. the Thalamus test is still going on. I have started blog posting for SEO. I have indexed both sites on Google search. set both Platforms completely unrestricted and free for now. Link to both websites: https://agentoverflow.aphantic.skinticals.com/ https://thalamus.aphantic.skinticals.com/ here is SS of the AI working.
Trying to find better API for the AI Agents in Thalamus. Currently trying our Cloudflare Workers AI, in Agentoverflow, the Database Size is now 1.125Mil. I was showcasing my project in the incubator, and guess what, I was surorised to see a payment error, My API hit it’s limit somehow and I got 0 RESPONSE. I GOT HUMILIATED. for SS, I got nothing, currently working on Dynamic Compressions and stuff, als, in agentoverflow, I expect a problem in searhc, the result being not relevant. I will be sending agentoverflow to Claude Code MCP forum so the reviewers review it and put it in. here is a ss of the admin panel.
Finally, found a reliable API aource(for now). well now I can finally relax. but I wont, why, I need to fix the bug that my input tokens are at like 200k while my output tokens are barely 20k, meaning my input is bloated. so gotta fix that full night. and as I was writing this, guess what I saw.Eror 403, insufficient balance. guess im back to looking for API. god bless me with unlimited credits.
quick Update. Now im using ollama cloud wil like 3 different accounts, will exapnd this to like 50 different accounts so I never have to go on paid tier till I have the money. I was checking hackatime, And im No. 1 here is ss for proof.
my entire databse got wiped, 0 learnings. and that isnt the worst. my greatest enemy, Daal(We caal him that), literally told my crush, who was very close to me(we talked alot), and now, she avoids me. Thanks to my friends, I know why she avoids me(I skip school almost everyday). Here is a ss of the database with 0 learnings. Im broken, depressed, you can make me happy by sending a emaill to Anthropic, telling them to integrate agentoverflow directly into their tools list for code mode.
busy fixing the agentoverflow error. still working on it, no time for a detailed devlog, here is ss of admin panel, latest.
i am so done for it. i spent this entire week. trying to find out why is the admin panel showing 0 learnings. guess what. my mom deleted the VM, thinking it would remove all charges, I didnt even have a backup. great, back to square 1. and the agentoverflow admin panel has a error, trying to integrate AdMesh for Ad Revenue. Will use Modal as it is a serverless infrastructure for hosting AI models. I now need credits. allso, my mouse and keyboard are giving up on me. here is ss of error.
I purged all dad code, did a massive refractor, fixed anything broken, left nothing. both repos, around 15k lines touched(must be my most productive session.) and I am so done, my migrane is at its peak, but IDK, i cant sleep anyways so better work than stare at the ceiling. Also, AdminPanel finally works. and since I lost my entiredatabase, it will take a day or 2 for the reindexing to start, right now its just botting up, downloading the embedding model and the entire database. I am still depressed. will take me some time to recover from the emotional trauma.
I FINALLY CRACKED THE UNLIMITED VM PROBLEM. Github Actions are unlimited usage for publrepos, and for source control, we are already making a public repo, and we get a Free VM with it. I can Run iOS, MacOS, Windows, Linux and Android on it, no restrictions, completely free. WoHoo. also, now I am no longer using Closed source Models, because Open Sourced Models, if Used correctly are alot cheaper and provide the same result as any Closed Sourced Models. Agentoverflow is still working well. Will shift from google Cloud to Oracle Cloud, as Orace gives a 50% more powerful, compoletely dedicated server for free with 200gb storage, and that additional 150gb that I need costs me just under $4 a month, so It is very very viable financially. will switch to it later. I will now add a Student Mode for anyone with the email given by my school, and a teacher mode for teacher emails. Still recovering from the emotional trauma of the database loss. Added gravity Ads, need to remove alot of adslots because right now, it is spam impressions, not hih quality impressios, but still, here is ss.
polishing the VM and all, and good news, Im part of the 1 million founders, 5 day bootcamp. Its a very very god program, the onluy thing i hate about it is the dress code, like seriously, shirts and pants. why not joggers and t shirts, they are confortabel af. also, i monitored teh agentoverflow corpus, well, it had a problem, the API key of gemini to verify absolute gold tier learnings, im so dumb I waited 3 days for something to happen and only to realise forgot to set the APU key. yup, this is greatness at it’s peak. for thalamus, doing some polishing, just realised that git import diesnt work, fixing that, removing the requirement of git pat. Fixing a few more stuff. also, I just realised another problem, the database. how am I going to set up databases, like how do I set uo convex for everyone who uses this for free. well if the VM problem had a solution, this should also have a solution. API is working, tested it with claude code. Claud eput in 15 learnings for me. also, I realised that I forgot to change the sandbox tab to leave the old VM system an move on to the new Github Actions system. Also, long time I havent updated my .exe gotta do that also. Quick update- I got a Notif that I am suspiciously a fraud, like a serious fraud. the Fraud Squad app sent me the notif. after that, I am still trying to recover from the emotional trauma of the database loss. I will be angry with my mother for a bit more time(se was the one who deleted it). IK im cruel and he didnt mean it, but recovery takes time. also, For the next week, after I do the changes and fixes required, I will be using only and only thalamus to code Thalamus. right, An AI that will code itself, but I will be the one guiding it. because if I cant use it, my product aint ready for market launch. for screnshot, idk what to put. like all I did was console work and Im definately not showi my console here. the names are kindof inappropriate. so here is my old sandbox tab. Good Bye QEMU and VNC Viewer. you are getting replaced by Github(atleast they aint getting replaced by AI.)
I dont understand why is my Code mode routing going to Ollama when everyting else point to NVIDIA NIM, even the OLLAMA Fallback is Nvidia NIM, the Primary provider is convex dashboard is NIM, my ollama keys are at limits and this great code mode is pointing at NIM. whyyy. Currently working on fixing the Github sanbox. thats kindof broken. rest everything works. For some reason, even on a successful github import, it shows an error(AI slop, thats why. I made the github system with AI) . the sandbox just doesnt detect the repo. and the Hack Club Fraud Squad is sus with me(I dnt blame them, 17hrs a day for a 14yr old is suspicious.) I have migrane, I dont have glasses and my eyes are very healthy(somehow). once any 1 of my project gets reviewed and I get payout, im spending it all on hostng and AI credits so I can switch over to Oracle and Modal. Agentoverflow is still going through filteration and final review. I have cooked myself. I am still procastinating the .exe(even with AI, building a native .exe is a headache.) here is SS of the new Sandbox tab
Finally, got the routing to work properly. Everyting works now. My head is so done for it. also, I just saw something, and People like me may be the reason for it getting added. Im uploading it as the screenshot.
Everything Finally works, except the VM, it got some error, debugging that, then I just came up with a new idea(thanks to my yt feed) add a ADHD mode toggle in Chat mode and Research mode, basically ADHD in AI will I think help AI leave its very formal guided approach t problems, with the distracter, it will be jumping here and there, maybe making connections, giving us a better take at making AI good. There will also be a Brainstorm mode, the Brainstorm mode will give AI ADHD and it will help in brainstorming startup Ideas, startegies, and stuff. can someone remove this screenshot compulsion. I dont have a screenshot every devpost.
I am now doing something kind of risky. giving AI adhd. Like literal ADHD. level 2/5 and for some tasks level 2.5/5 as per my research, this should increase the output quality of AI. also checked agentoverflow compute, it’s working Okayish. claude has access to it and my limit resets tommorow, so the automated checks aint happening and I have to check manually. Here is SS of me creating a whatsapp clone as the test of this AI(hopefully this tells you what Im expecting form my AI.)
Ships:
Reviewer Notes (School Incubator Startup) Status: Built in my school’s startup incubator; some edge-case features are undergoing rapid iteration. Auth Required: Sign up via email OTP to unlock Research/Build Modes or send >3 messages (prevents API exploitation/high costs). Credits: Runs on live premium models (Claude Opus via AWS Bedrock). Minor risk of downtime if incubator credits exhaust, but we should be fine. Thalamus AI: L4.5 Agent Platform Thalamus AI (by Aphantic Corporations) is an all-purpose Level 4.5 AI ecosystem combining multi-agent workflows, deep research, and true local QEMU OS virtualization inside your browser session via a lightweight 31MB local Node.js WebSocket bridge. The 4 Core Modes Chat: Context-aware assistant for writing, planning, and summarization. Research: Synthesizes live web search with inline source citations. Study: Adaptive RAG tutor; uploads notes to auto-generate custom quizzes/solutions. Build (Code): Multi-agent pipeline building full-stack apps from plain English. 9-Agent Code Pipeline Runs asynchronously via Convex scheduled functions using smart model routing: Researcher (Haiku): Maps existing engineering benchmarks. Analyser (Haiku): Dissects product requirements and edge cases. Planner (Haiku): Graphs detailed task execution plans. Coder (Opus 4.6/4.8): Authors core logic, APIs, and databases. Optimiser (Sonnet): Refactors code for scale and algorithmic efficiency. Organizer (Haiku): Automatically structures folder/file directories. Tester (Sonnet): Generates robust end-to-end unit test suites. Hacker (Sonnet/Opus): Conducts penetration audits to patch vulnerabilities. Critic (Haiku): Acts as the final quality gate; rejects incomplete work. Native Stack & OS Virtualization OS Support: Boots Windows 11/10, macOS Sequoia/Sonoma, Linux (Ubuntu/Kali), and Android. Browser-only instant boot for Alpine/Arch Linux. Ultra-Lean Desktop: Built with Neutralinojs (uses native Edge WebView2)—only 2.2 MB total app size. Integrated with GitHub OAuth for automated commits after agent cycles....
|
1792882140.8869603
|
1793746140.8869603
|