|
32967
|
Dionysiaca by @anujkishan918 | Stardance
|
anujkishan918
|
0
|
74.07
|
0
|
[]
|
[{"post_id": 34155, "ts": "202 [{"post_id": 34155, "ts": "2026-07-24T17:40:51Z", "hours": 7.9667, "text": "Final Devlog(hopefully :) ):This project taught me a lot more than just how to use Godot.When I started, I thought game development was mostly about writing code and making mechanics work. Somewhere along the way, I realized it’s just as much about knowing when to simplify, when to step away, and when to stop chasing the “perfect” feature.There were days when everything clicked and I made more progress than I expected. There were also days where I spent hours fixing one tiny bug, hit burnout, or even forgot to post a devlog and paid the price for it. Looking back, those moments are just as much a part of this project as the finished features.I’m really happy with where the game is now. It’s gone from a few scattered ideas to something you can actually play, with combat, enemies, checkpoints, UI, and a world that finally feels alive. I even managed to hide a little Easter egg that I hope someone smiles at when they find it.This definitely isn’t the end of the project,just the end of this chapter. More than anything, I’m a lot more confident in Godot and in my ability to build something from scratch."}, {"post_id": 33597, "ts": "2026-07-23T17:29:46Z", "hours": 15.4361, "text": "Devlog #8I completely forgot to post yesterday’s devlog, and that mistake ended up costing me 5 hours. It was definitely frustrating, but I guess that’s part of sticking to a challenge like this—building the game is only half the job, documenting the journey matters too.On the development side, I made good progress despite the setback. I refactored parts of the enemy AI state machine to make transitions between Patrol, Chase, and Attack states more reliable, reducing animation glitches during combat. I also fine-tuned hitbox timings and attack cooldowns so enemy encounters feel more responsive and fair. Along the way, I fixed a few collision issues where enemies could occasionally get stuck against TileMap edges.To keep the project maintainable, I reorganized several scenes and scripts, reducing duplicate logic and making it easier to add future enemy types. It wasn’t the most exciting work, but it made the codebase much cleaner.Losing five hours hurt, but the game itself is in a much stronger state than it was a few days ago. Every bug fixed and every system cleaned up is another step toward the first playable milestone."}, {"post_id": 30265, "ts": "2026-07-17T18:13:14Z", "hours": 8.0667, "text": "Devlog #7: Today felt like a turning point after the slowdown earlier this week. With the scope now under control, I was able to focus on making meaningful progress instead of constantly jumping between unfinished features.I expanded Level 1 by adding new combat encounters and refining enemy placement to create a smoother difficulty curve. I also continued polishing the combat system by improving attack responsiveness, animation transitions, and enemy interactions, making each fight feel more satisfying and consistent.Beyond gameplay, I cleaned up parts of the project structure, removing temporary code and reorganizing scenes to make future development easier. It’s one of those changes players will never notice, but it makes a huge difference when building and maintaining the game.For the first time, the game feels like a complete vertical slice rather than a collection of separate systems. There’s still plenty to build, but the foundation is solid, and I’m getting closer to the first milestone with every session."}, {"post_id": 29595, "ts": "2026-07-16T17:10:18Z", "hours": 6.8856, "text": "Devlog #6The past two days have been rough. I hit a pretty major burnout, which I guess is just part of being a maker. I got caught up trying to implement too many features at once, and instead of making steady progress, I ended up overwhelming myself. To top it off, I lost my 17-day streak, which was a little frustrating.Oddly enough, watching Project Hail Mary helped me reset my mindset. The optimism and overall vibe of the movie reminded me why I enjoy building things in the first place. Sometimes stepping away for a bit is exactly what you need.With a clearer head, I decided to scale back a few features and focus on what actually matters for the first milestone. That decision has made development feel much more manageable, and I genuinely think the game is now very close to reaching that milestone.I also snuck in a small secret Easter egg that I hope players discover and enjoy. It’s a tiny detail, but I think little surprises like that make a game feel more memorable."}, {"post_id": 27774, "ts": "2026-07-13T16:59:42Z", "hours": 1.2333, "text": "Devlog #5 Not a huge development day, but I made some quality-of-life improvements. I designed the game’s starting UI and also refined the existing boss fight interface to make it cleaner and more consistent with the overall visual style. Small UI changes like these go a long way in making the game feel more polished.Tomorrow, I’ll get back to gameplay development and continue working on Level 1."}, {"post_id": 27275, "ts": "2026-07-12T18:28:19Z", "hours": 8.6497, "text": "Devlog #4Today’s work was focused on refining the first combat encounter rather than adding lots of new content. I expanded the enemy AI by improving its patrol and chase behavior, making transitions between states much smoother and more predictable. I also spent time adjusting hitboxes, attack ranges, and movement speeds so fights feel fair while still requiring good timing from the player.I added the first environmental obstacle to the level and made sure it integrates correctly with the existing TileMap and collision layers. Most of the day, however, was spent debugging. Small issues like animation interruptions, collision edge cases, and state transitions took longer than expected, but fixing them has made the gameplay feel noticeably more polished.While Level 1 is still a work in progress, each iteration is making the core gameplay loop more enjoyable. My next goal is to add another enemy type and continue building out the encounters so the level gradually introduces more challenging combat scenarios."}, {"post_id": 26694, "ts": "2026-07-11T17:12:22Z", "hours": 9.2842, "text": "Today was all about bringing Level 1 to life. After finishing the layout yesterday, I shifted my focus from building the environment to implementing gameplay. I integrated the first enemy into the level, created its AI state machine, and connected it to the player’s combat system so encounters now feel like actual gameplay instead of isolated mechanics.A significant part of the day went into tuning collision detection, attack timing, and animation synchronization. Since the player now has both light and charged heavy attacks, I had to ensure the enemy reacts correctly to different attack windows without breaking the animation flow. I also refined the checkpoint logic so the shrine is now properly integrated into the gameplay loop.What I’m happiest about is that the project is finally starting to feel like a game rather than a collection of individual systems. The combat, movement, checkpoint, and level all exist together in a single playable experience. Tomorrow I’ll focus on expanding enemy behavior, adding more encounters, and introducing environmental obstacles to improve the pacing and challenge of Level 1."}, {"post_id": 26104, "ts": "2026-07-10T18:06:59Z", "hours": 8.6167, "text": "Today was one of my biggest learning days with Godot so far. I spent most of my time understanding how the engine handles player mechanics, animation state transitions, and level construction, and I am very happy with the progress I made.The achievement I’m most proud of today was creating and integrating a completely new charged heavy attack for the player. Instead of simply reusing an existing animation, I designed a custom pixel-art sprite sheet that matches the style of the rest of the character animations and successfully integrated it into my animation state machine. I updated the player controller to support a dedicated HEAVY_ATTACK state with its own animation timing, input handling, and state transitions, ensuring it behaves independently from the normal attack combo while remaining compatible with the rest of the combat system.On the level design side, I completed the entire layout for Level 1 using Godot’s TileMap system. This involved constructing the full traversable floor, shaping the terrain, and ensuring the level can support future gameplay systems such as combat encounters, checkpoints, and environmental hazards. Moving to a TileMap-based workflow has already made level creation significantly more efficient and scalable compared to manually placing collision objects.Although the level is not yet populated with enemies, I implemented the first checkpoint by placing a shrine/statue that will serve as the player’s respawn point. Since checkpoint systems are a core part of Souls-inspired games, getting this foundation in place was an important milestone for the overall gameplay loop.Tomorrow’s goal is to begin populating the level with enemy AI, obstacles, and combat encounters. Once those systems are in place, Level 1 will become fully playable from start to finish, allowing me to start balancing exploration, difficulty progression, and combat pacing."}, {"post_id": 25350, "ts": "2026-07-09T16:12:56Z", "hours": 7.9333, "text": "Devlog #1 — Finally Starting My First Godot GameI’ve finally started working on my first 2D pixel art game with Godot. The idea for this game has been in my head for a long time. It’s based on a part of Greek mythology that doesn’t get talked about very often, and I’ve wanted to turn that story into a game for years. But before I could even think about making that game properly, I knew I first had to learn how to make games.Most of my time this week actually wasn’t spent inside Godot. It went into all the background work that people usually don’t see. I spent hours finding, generating, and organizing pixel art assets, making sure they worked well together and were all roughly the same style and scale. I sorted everything into folders for characters, enemies, tilesets, UI, backgrounds, and animations so the project wouldn’t become a mess later on.Learning Godot was honestly harder than I expected. I watched quite a few YouTube tutorials and read different guides, but I always felt like I was just copying what someone else was doing instead of understanding it. What finally helped was Godot’s own Learn to Code with Godot course/app and this one is really hands on . It starts with the basics and slowly introduces GDScript, nodes, scenes, signals, variables, loops, and functions. After working through it, I finally felt like I understood why things worked instead of just memorizing code.Once I felt comfortable enough, I created the project, imported all of my assets, set up the project for pixel-perfect rendering, and gave it a temporary (very goofy) name that will definitely change later. I also built a simple title screen and a basic HUD. One of my favorite moments was getting the menus fully navigable with just the keyboard—it made the game feel like an actual game instead of just a collection of sprites on the screen.This first milestone wasn’t really about gameplay. It was about building a solid starting point and getting familiar with Godot’s workflow. Even though there isn’t much to “play” yet, I learned a lot about how scenes are structured, how nodes fit together, and how everything in Godot connects.Next, I’m moving on to the second milestone, where the real fun begins. The plan is to build the first playable side-scrolling level inspired by games like Dead Cells. I’ll be focusing on the core mechanics player movement, jumping, collisions, animations, camera movement, enemy behavior, combat, and interactions. If I can get those systems feeling good, I’ll have a strong foundation to start building the rest of the game on."}]...
|
Title: Dionysiaca by @anujkishan918 | Stardance
D Title: Dionysiaca by @anujkishan918 | Stardance
Devlogs:
Devlog #1 — Finally Starting My First Godot GameI’ve finally started working on my first 2D pixel art game with Godot. The idea for this game has been in my head for a long time. It’s based on a part of Greek mythology that doesn’t get talked about very often, and I’ve wanted to turn that story into a game for years. But before I could even think about making that game properly, I knew I first had to learn how to make games.Most of my time this week actually wasn’t spent inside Godot. It went into all the background work that people usually don’t see. I spent hours finding, generating, and organizing pixel art assets, making sure they worked well together and were all roughly the same style and scale. I sorted everything into folders for characters, enemies, tilesets, UI, backgrounds, and animations so the project wouldn’t become a mess later on.Learning Godot was honestly harder than I expected. I watched quite a few YouTube tutorials and read different guides, but I always felt like I was just copying what someone else was doing instead of understanding it. What finally helped was Godot’s own Learn to Code with Godot course/app and this one is really hands on . It starts with the basics and slowly introduces GDScript, nodes, scenes, signals, variables, loops, and functions. After working through it, I finally felt like I understood why things worked instead of just memorizing code.Once I felt comfortable enough, I created the project, imported all of my assets, set up the project for pixel-perfect rendering, and gave it a temporary (very goofy) name that will definitely change later. I also built a simple title screen and a basic HUD. One of my favorite moments was getting the menus fully navigable with just the keyboard—it made the game feel like an actual game instead of just a collection of sprites on the screen.This first milestone wasn’t really about gameplay. It was about building a solid starting point and getting familiar with Godot’s workflow. Even though there isn’t much to “play” yet, I learned a lot about how scenes are structured, how nodes fit together, and how everything in Godot connects.Next, I’m moving on to the second milestone, where the real fun begins. The plan is to build the first playable side-scrolling level inspired by games like Dead Cells. I’ll be focusing on the core mechanics player movement, jumping, collisions, animations, camera movement, enemy behavior, combat, and interactions. If I can get those systems feeling good, I’ll have a strong foundation to start building the rest of the game on.
Today was one of my biggest learning days with Godot so far. I spent most of my time understanding how the engine handles player mechanics, animation state transitions, and level construction, and I am very happy with the progress I made.The achievement I’m most proud of today was creating and integrating a completely new charged heavy attack for the player. Instead of simply reusing an existing animation, I designed a custom pixel-art sprite sheet that matches the style of the rest of the character animations and successfully integrated it into my animation state machine. I updated the player controller to support a dedicated HEAVY_ATTACK state with its own animation timing, input handling, and state transitions, ensuring it behaves independently from the normal attack combo while remaining compatible with the rest of the combat system.On the level design side, I completed the entire layout for Level 1 using Godot’s TileMap system. This involved constructing the full traversable floor, shaping the terrain, and ensuring the level can support future gameplay systems such as combat encounters, checkpoints, and environmental hazards. Moving to a TileMap-based workflow has already made level creation significantly more efficient and scalable compared to manually placing collision objects.Although the level is not yet populated with enemies, I implemented the first checkpoint by placing a shrine/statue that will serve as the player’s respawn point. Since checkpoint systems are a core part of Souls-inspired games, getting this foundation in place was an important milestone for the overall gameplay loop.Tomorrow’s goal is to begin populating the level with enemy AI, obstacles, and combat encounters. Once those systems are in place, Level 1 will become fully playable from start to finish, allowing me to start balancing exploration, difficulty progression, and combat pacing.
Today was all about bringing Level 1 to life. After finishing the layout yesterday, I shifted my focus from building the environment to implementing gameplay. I integrated the first enemy into the level, created its AI state machine, and connected it to the player’s combat system so encounters now feel like actual gameplay instead of isolated mechanics.A significant part of the day went into tuning collision detection, attack timing, and animation synchronization. Since the player now has both light and charged heavy attacks, I had to ensure the enemy reacts correctly to different attack windows without breaking the animation flow. I also refined the checkpoint logic so the shrine is now properly integrated into the gameplay loop.What I’m happiest about is that the project is finally starting to feel like a game rather than a collection of individual systems. The combat, movement, checkpoint, and level all exist together in a single playable experience. Tomorrow I’ll focus on expanding enemy behavior, adding more encounters, and introducing environmental obstacles to improve the pacing and challenge of Level 1.
Devlog #4Today’s work was focused on refining the first combat encounter rather than adding lots of new content. I expanded the enemy AI by improving its patrol and chase behavior, making transitions between states much smoother and more predictable. I also spent time adjusting hitboxes, attack ranges, and movement speeds so fights feel fair while still requiring good timing from the player.I added the first environmental obstacle to the level and made sure it integrates correctly with the existing TileMap and collision layers. Most of the day, however, was spent debugging. Small issues like animation interruptions, collision edge cases, and state transitions took longer than expected, but fixing them has made the gameplay feel noticeably more polished.While Level 1 is still a work in progress, each iteration is making the core gameplay loop more enjoyable. My next goal is to add another enemy type and continue building out the encounters so the level gradually introduces more challenging combat scenarios.
Devlog #5 Not a huge development day, but I made some quality-of-life improvements. I designed the game’s starting UI and also refined the existing boss fight interface to make it cleaner and more consistent with the overall visual style. Small UI changes like these go a long way in making the game feel more polished.Tomorrow, I’ll get back to gameplay development and continue working on Level 1.
Devlog #6The past two days have been rough. I hit a pretty major burnout, which I guess is just part of being a maker. I got caught up trying to implement too many features at once, and instead of making steady progress, I ended up overwhelming myself. To top it off, I lost my 17-day streak, which was a little frustrating.Oddly enough, watching Project Hail Mary helped me reset my mindset. The optimism and overall vibe of the movie reminded me why I enjoy building things in the first place. Sometimes stepping away for a bit is exactly what you need.With a clearer head, I decided to scale back a few features and focus on what actually matters for the first milestone. That decision has made development feel much more manageable, and I genuinely think the game is now very close to reaching that milestone.I also snuck in a small secret Easter egg that I hope players discover and enjoy. It’s a tiny detail, but I think little surprises like that make a game feel more memorable.
Devlog #7: Today felt like a turning point after the slowdown earlier this week. With the scope now under control, I was able to focus on making meaningful progress instead of constantly jumping between unfinished features.I expanded Level 1 by adding new combat encounters and refining enemy placement to create a smoother difficulty curve. I also continued polishing the combat system by improving attack responsiveness, animation transitions, and enemy interactions, making each fight feel more satisfying and consistent.Beyond gameplay, I cleaned up parts of the project structure, removing temporary code and reorganizing scenes to make future development easier. It’s one of those changes players will never notice, but it makes a huge difference when building and maintaining the game.For the first time, the game feels like a complete vertical slice rather than a collection of separate systems. There’s still plenty to build, but the foundation is solid, and I’m getting closer to the first milestone with every session.
Devlog #8I completely forgot to post yesterday’s devlog, and that mistake ended up costing me 5 hours. It was definitely frustrating, but I guess that’s part of sticking to a challenge like this—building the game is only half the job, documenting the journey matters too.On the development side, I made good progress despite the setback. I refactored parts of the enemy AI state machine to make transitions between Patrol, Chase, and Attack states more reliable, reducing animation glitches during combat. I also fine-tuned hitbox timings and attack cooldowns so enemy encounters feel more responsive and fair. Along the way, I fixed a few collision issues where enemies could occasionally get stuck against TileMap edges.To keep the project maintainable, I reorganized several scenes and scripts, reducing duplicate logic and making it easier to add future enemy types. It wasn’t the most exciting work, but it made the codebase much cleaner.Losing five hours hurt, but the game itself is in a much stronger state than it was a few days ago. Every bug fixed and every system cleaned up is another step toward the first playable milestone.
Final Devlog(hopefully :) ):This project taught me a lot more than just how to use Godot.When I started, I thought game development was mostly about writing code and making mechanics work. Somewhere along the way, I realized it’s just as much about knowing when to simplify, when to step away, and when to stop chasing the “perfect” feature.There were days when everything clicked and I made more progress than I expected. There were also days where I spent hours fixing one tiny bug, hit burnout, or even forgot to post a devlog and paid the price for it. Looking back, those moments are just as much a part of this project as the finished features.I’m really happy with where the game is now. It’s gone from a few scattered ideas to something you can actually play, with combat, enemies, checkpoints, UI, and a world that finally feels alive. I even managed to hide a little Easter egg that I hope someone smiles at when they find it.This definitely isn’t the end of the project,just the end of this chapter. More than anything, I’m a lot more confident in Godot and in my ability to build something from scratch....
|
1792940076.0374951
|
1793804076.0374951
|
|
23725
|
ForgeMind by @FazinAhamed | Stardance
|
FazinAhamed
|
0
|
29.0
|
0
|
[]
|
[{"post_id": 29729, "ts": "202 [{"post_id": 29729, "ts": "2026-07-16T20:23:03Z", "hours": 8.0, "text": "past 8hrs was disastrous like I got inflicted a debuff called Sinusitis and a fever prolly cuz ive been avoiding sleep for a long time. Anyways, I kept working on ForgeMind source storage and archive tracking. Archives now have a proper identity beyond just their folder path, so each run can point back to the exact source state that produced an index. In other words, reproducibility is slowly moving from “Source: Trust me bro🥀🥀” to something that can actually be checked and enforced so it is really stronk proof. I also worked on refresh behaviour. When one file changes, ForgeMind can now update it without treating the entire archive like a completely new corpus. It sounds pretty useless, but it makes repeated runs much faster and far more practical so that you dont waste compute and power. retrieval also has a cleaner path back to the stored source identity. This matters a lot because an answer is only useful when its citations can still be checked against the exact archive version they came from. There were a few errors, random blockers, and my brain was running on half power and tissue box for most of the day, but the foundation is getting stronger. Not the flashiest update and fabulous cool shi but definitely a off-camera grinding sesh worthy improvement, I think. Raters if u see this pls i need this PS:. The new comments and extra attention to detail like explaining each thing is cuz of tejas so hope its enough Edit: Also i will be commiting a change today or tmrw cuz i got this to work. Srry i havent been committing, just got hella stressed cuz stuff wasnt working out during tests."}, {"post_id": 29407, "ts": "2026-07-16T10:34:02Z", "hours": 4.0, "text": "Spent the last 4 hours on source storage, making the archive less like a temporary input and more like something ForgeMind can rely on across multiple runs. The focus was keeping source identity stable and also paths, hashes, refresh behavior, and stored metadata now have a clearer place to live instead of being reconstructed unplanned and temporarily. That matters because retrieval and verification only mean something if they point back to the exact source state that produced them. So nothing insane and cool today, but it clears an important gap, the system is now starting to remember what it saw, not just process it once. So it is a win for me even though it took quite a while"}, {"post_id": 29320, "ts": "2026-07-16T05:18:28Z", "hours": 1.0, "text": "Good Morning, Woke up had some plain black coffee and hit the roads again. Today I finished some stuff like converting the first scale result into a deterministic gate instead of leaving it as a one-off demo also now the archive generator uses a fixed seed, the tokenizer and model artifacts are hash pinned, and the profiler records ingestion, refresh, compression, and retrieval measurements in machine readable output. Essentially just implying, reproducing a result is now part of the feature. Next i will probably get to the source storage and ye thats it for this devlog PS. Pls dont judge me for using notepad, its just more convenient sometimes even for markdown files"}, {"post_id": 29090, "ts": "2026-07-15T19:05:30Z", "hours": 6.0, "text": "ForgeMind Devlop - First Real bit of Proof ig and The First Proper devlog The first ever milestone for ForgeMind is working and tested on end to end consumer and corporate hardware I am running a quantized Qwen3-4B through llama.cpp on an RTX 3060 with 12gigs of VRAM. ForgeMind was able to ingest an archive containing 4.06M Qwen tokens, divided it into searchable chunks of 1155 and retrieved info while keeping the prompt bound to around 15.5k tokens Disclaimer - This is virtual long context, not a claim that the model directly attends to four million tokens. The entire archive remains outside the attention. ForgeMind just retrieves relevant passages, rehydrates the exact source text and verifies citations using source hashes and line spans The TokenForge subpart produced quite measurable and valuable efficiency gains, It reversibly compressed eligible passages by an average or calculated median of around ~14% , with a 100% exact restoration. The Cold Ingests took 100.1 seconds, while refreshing one changed file at equivalent scale took only 1.04 s and reembedded only 5 affected chunks . The automated gates and tests and benchmarks now passes 55 with zero errors or failures. I also did 10 offline / local checks which also completed successfully, during which the citations remained valid and the runtime hesitates when evidence is unavailable. The 4.07M token test proves scalable ingestion, reversible compression and incremental refresh and also the retrievals. Thats all for this. next i will turn this proof into a controlled comparison b/w raw context, hybrid rag and vector rag. Thats all for this."}, {"post_id": 28782, "ts": "2026-07-15T08:48:38Z", "hours": 5.0, "text": "I havent slept since yesterday, my eyes are about to die on me but i reached a breakthroughhh!!!! finally made this CPU Native and might be able to achieve one more thing which i will disclose next devlog. This will be huge for Local LLM users. This is possible using something called ThinkingCaps and ContextFabric. These methods will allow for actually usable fast 9B CPU native model inferencing with like enough context length and Symbolic Reasoning and Coding capabilities and verification stuff Also made a Web UI for Investigating and testing out smaller quantities of test runs which is supposed to run on Local. I will probably send the first milestone report for next devlog and also start working on a Demo URL that will be hosted on Modal"}, {"post_id": 28508, "ts": "2026-07-14T20:04:25Z", "hours": 2.0, "text": "Finally got 4M tokens to fit inside a 32k context windowwww!!!! Les gooo and also hope france wins"}, {"post_id": 28414, "ts": "2026-07-14T17:30:19Z", "hours": 2.0, "text": "Made some tests and got an Abstract ready"}, {"post_id": 28320, "ts": "2026-07-14T14:33:08Z", "hours": 1.0, "text": "Divided my work into 8 Milestones and confirmed that i have basically finished till m1.5"}, {"post_id": 28276, "ts": "2026-07-14T13:00:56Z", "hours": 0.0, "text": "Just started the project. Its essentially an AI Model Cost and Performance Optimizer thing"}]...
|
Title: ForgeMind by @FazinAhamed | Stardance
Devl Title: ForgeMind by @FazinAhamed | Stardance
Devlogs:
Just started the project. Its essentially an AI Model Cost and Performance Optimizer thing
Divided my work into 8 Milestones and confirmed that i have basically finished till m1.5
Made some tests and got an Abstract ready
Finally got 4M tokens to fit inside a 32k context windowwww!!!! Les gooo and also hope france wins
I havent slept since yesterday, my eyes are about to die on me but i reached a breakthroughhh!!!! finally made this CPU Native and might be able to achieve one more thing which i will disclose next devlog. This will be huge for Local LLM users. This is possible using something called ThinkingCaps and ContextFabric. These methods will allow for actually usable fast 9B CPU native model inferencing with like enough context length and Symbolic Reasoning and Coding capabilities and verification stuff Also made a Web UI for Investigating and testing out smaller quantities of test runs which is supposed to run on Local. I will probably send the first milestone report for next devlog and also start working on a Demo URL that will be hosted on Modal
ForgeMind Devlop - First Real bit of Proof ig and The First Proper devlog The first ever milestone for ForgeMind is working and tested on end to end consumer and corporate hardware I am running a quantized Qwen3-4B through llama.cpp on an RTX 3060 with 12gigs of VRAM. ForgeMind was able to ingest an archive containing 4.06M Qwen tokens, divided it into searchable chunks of 1155 and retrieved info while keeping the prompt bound to around 15.5k tokens Disclaimer - This is virtual long context, not a claim that the model directly attends to four million tokens. The entire archive remains outside the attention. ForgeMind just retrieves relevant passages, rehydrates the exact source text and verifies citations using source hashes and line spans The TokenForge subpart produced quite measurable and valuable efficiency gains, It reversibly compressed eligible passages by an average or calculated median of around ~14% , with a 100% exact restoration. The Cold Ingests took 100.1 seconds, while refreshing one changed file at equivalent scale took only 1.04 s and reembedded only 5 affected chunks . The automated gates and tests and benchmarks now passes 55 with zero errors or failures. I also did 10 offline / local checks which also completed successfully, during which the citations remained valid and the runtime hesitates when evidence is unavailable. The 4.07M token test proves scalable ingestion, reversible compression and incremental refresh and also the retrievals. Thats all for this. next i will turn this proof into a controlled comparison b/w raw context, hybrid rag and vector rag. Thats all for this.
Good Morning, Woke up had some plain black coffee and hit the roads again. Today I finished some stuff like converting the first scale result into a deterministic gate instead of leaving it as a one-off demo also now the archive generator uses a fixed seed, the tokenizer and model artifacts are hash pinned, and the profiler records ingestion, refresh, compression, and retrieval measurements in machine readable output. Essentially just implying, reproducing a result is now part of the feature. Next i will probably get to the source storage and ye thats it for this devlog PS. Pls dont judge me for using notepad, its just more convenient sometimes even for markdown files
Spent the last 4 hours on source storage, making the archive less like a temporary input and more like something ForgeMind can rely on across multiple runs. The focus was keeping source identity stable and also paths, hashes, refresh behavior, and stored metadata now have a clearer place to live instead of being reconstructed unplanned and temporarily. That matters because retrieval and verification only mean something if they point back to the exact source state that produced them. So nothing insane and cool today, but it clears an important gap, the system is now starting to remember what it saw, not just process it once. So it is a win for me even though it took quite a while
past 8hrs was disastrous like I got inflicted a debuff called Sinusitis and a fever prolly cuz ive been avoiding sleep for a long time. Anyways, I kept working on ForgeMind source storage and archive tracking. Archives now have a proper identity beyond just their folder path, so each run can point back to the exact source state that produced an index. In other words, reproducibility is slowly moving from “Source: Trust me bro🥀🥀” to something that can actually be checked and enforced so it is really stronk proof. I also worked on refresh behaviour. When one file changes, ForgeMind can now update it without treating the entire archive like a completely new corpus. It sounds pretty useless, but it makes repeated runs much faster and far more practical so that you dont waste compute and power. retrieval also has a cleaner path back to the stored source identity. This matters a lot because an answer is only useful when its citations can still be checked against the exact archive version they came from. There were a few errors, random blockers, and my brain was running on half power and tissue box for most of the day, but the foundation is getting stronger. Not the flashiest update and fabulous cool shi but definitely a off-camera grinding sesh worthy improvement, I think. Raters if u see this pls i need this PS:. The new comments and extra attention to detail like explaining each thing is cuz of tejas so hope its enough Edit: Also i will be commiting a change today or tmrw cuz i got this to work. Srry i havent been committing, just got hella stressed cuz stuff wasnt working out during tests....
|
1792940070.243705
|
1793804070.243705
|