Trail - Public Entries

Single photo to pixel-aligned 3D. Each vertex corresponds to a real pixel - no guessing. SIGGRAPH 2026. Open source. https://github.com/TencentARC/Pixal3D

#3d-generation #computer-vision #deep-learning #generative-ai #image-to-3d #neural-rendering #pytorch #siggraph

GitHub - TencentARC/Pixal3D: [SIGGRAPH 2026] Pixal3D: Pixel-Aligned 3D Generation from Images

[SIGGRAPH 2026] Pixal3D: Pixel-Aligned 3D Generation from Images - TencentARC/Pixal3D

GitHub

1

Kibotu · 1d ago

Forgejo: self-hosted Git hosting that just works. Forked from Gitea, now its own thing. Lightweight, AGPLv3+, no surprises. https://codeberg.org/forgejo/forgejo

#codeberg #devops #forgejo #git #gitea #open-source #self-hosted #version-control

forgejo

Beyond coding. We forge.

Codeberg.org

6

Kibotu · 2d ago

Building with Claude? May 6 in SF: keynotes, workshops, live demos from Anthropic. https://claude.com/code-with-claude/san-francisco?utm_campaign=40539426-2026.05.06-Event-Code-with-Claude-Livestream

#ai-assisted-development #anthropic #code-with-claude #event #keynotes #live-demos #san-francisco #workshops

Code with Claude San Francisco — May 6, 2026

Keynotes, workshops, and live demos with the Anthropic team. Explore the future of AI-assisted development in San Francisco.

Claude

9

Kibotu · 2d ago

Most Docker dashboards guess at memory. ContainerFlow measures it. Real-time metrics, Discord alerts, config suggestions. AGPL-3.0. https://www.reddit.com/r/coolgithubprojects/comments/1ta8kak/typescript_containerflow_realtime_docker/

#dashboard #devops #docker #monitoring #open-source #typescript

[TypeScript] ContainerFlow - Real-time Docker dashboard with accurate memory monitoring, Discord alerts, and config recommendations

94 votes, 17 comments. Open-source real-time Docker dashboard I just released (AGPL-3.0). Visualizes all your containers and their connections in an…

13

Kibotu · 4d ago

Fused KV dequant into flash attention's inner loop. 82+ tok/s lossless 4-bit at 262K on a 4090. Clean hack, nobody shipped it sooner. https://indrasmirror.au/blog-mtp-shared-tensors-200k

#flash-attention #gpu-inference #kv-cache #llm-optimization #machine-learning #quantization #rtx-4090 #transformer

Fused TBQ4 Flash Attention: 80-87 tok/s with Lossless 4-bit KV at 262K Context

We fused quantized-KV dequant directly into the flash attention inner loop — something nobody else has done. 82+ tok/s on RTX 4090 with 4.25 bpv lossless KV compression.

Indra's Mirror

13

Kibotu · 4d ago

AI uplift on old tasks, new tasks, and in value aren't interchangeable. Task substitution is why. https://metr.org/blog/2026-05-08-task-substitution-and-uplift/

#ai-impact #ai-metrics #ai-research #ai-uplift #artificial-intelligence #metr #task-performance #task-substitution

Task Substitution and Uplift

We distinguish three measures of AI uplift -- on old tasks, on new tasks, and in value -- and show that task substitution can cause these to diverge substantially.

METR

16

Kibotu · 5d ago

Agent harness for Claude Code, Codex, Opencode, Cursor. Skills, instincts, memory, security — all performance-focused, research-first, and refreshingly free of nonsense. https://github.com/affaan-m/everything-claude-code

#agent-harness #ai-agents #claude-code #codex #cursor #developer-tools #opencode #performance-optimization

GitHub - affaan-m/everything-claude-code: The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond. - affaan-m/everything-cla...

GitHub

10

Kibotu · May 7

Test runner for AI agent skills. Cuts through the demos to measure what agents actually do. https://github.com/darkrishabh/agent-skills-eval

#agent-skills #ai #benchmark #evaluation #github #test-runner #testing

GitHub - darkrishabh/agent-skills-eval: A test runner for agentskills.io-style AI agent skills

A test runner for agentskills.io-style AI agent skills - darkrishabh/agent-skills-eval

GitHub

17

Kibotu · May 6

Qwen 3.6 at 35B parameters, quantized to GGUF by Unsloth. Handy to browse the repo layout before pulling down a multi-gigabyte model. https://hfviewer.com/unsloth/Qwen3.6-35B-A3B-GGUF

#artificial-intelligence #gguf #hugging-face #large-language-model #natural-language-processing #qwen #transformer #unsloth

unsloth/Qwen3.6-35B-A3B-GGUF • Hugging Face Viewer

Open unsloth/Qwen3.6-35B-A3B-GGUF in the interactive Hugging Face Viewer.

hfviewer.com

11

Kibotu · May 6

A satirical GitHub contribution graph. Red squares track platform outages instead of commits. Honestly, some repos would look identical. https://red-squares.cian.lol/

#contribution-graph #data-visualization #developer-humor #github #github-outages #open-source #outage-tracker #satirical

Red Squares — the GitHub outage graph

A satirical contribution graph: red squares track GitHub.com platform outages instead of green squares tracking commits.

red-squares.cian.lol

19

Kibotu · May 6

Zero-shot roguelite from one spec prompt, 20 minutes flat. Surprisingly playable. https://github.com/kibotu/cozylike

#ai-generated #game-development #prompt-engineering #roguelite #zero-shot

GitHub - kibotu/cozylike: roguelite generated zero-shot by one spec prompt in 20min.

roguelite generated zero-shot by one spec prompt in 20min. - kibotu/cozylike

GitHub

14

Kibotu · May 6

Turns out code generation is where AI actually belongs. If you're still treating it as an IDE plugin instead of your primary interface, you're optimizing for friction. Practical walkthrough of an AI-first workflow. https://www.chrismdp.com/coding-with-ai/

#ai-assistant #ai-coding #ai-programming #ai-workflow #coding-tools #cursor-ide #github-copilot #software-development

How I Use AI to Code

If you are still tied to your IDE, whether Cursor or Copilot, you are working a year behind. Coding turned out to be AI’s home territory. The best tooling ha...

Chris Parsons

11

Kibotu · May 6

Chrome DevTools speaks MCP now. Coding agents get direct access to inspect, debug, and profile without the browser automation layer. The sort of thing that just works. https://github.com/ChromeDevTools/chrome-devtools-mcp

#ai-agents #automation #browser-devtools #chrome-devtools #coding-agents #developer-tools #mcp

GitHub - ChromeDevTools/chrome-devtools-mcp: Chrome DevTools for coding agents

Chrome DevTools for coding agents. Contribute to ChromeDevTools/chrome-devtools-mcp development by creating an account on GitHub.

GitHub

13

Kibotu · May 6

Karpathy’s self-maintaining knowledge base implementation guide uses no DB or Obsidian plugins. Just markdown, an agent, three folders, a schema. Worth the read: https://www.reddit.com/r/AIAgentsInAction/s/Gi7FtSBC7V

#ai-agents #implementation-guide #knowledge-base #markdown #self-maintaining

Karpathy’s knowledge base that maintains itself. Implementation Guide. it has three folders and one schema file.

130 votes, 17 comments. I run a knowledge base that organises itself. No database. No Obsidian plugins. Just markdown files and an agent that…

13

Kibotu · May 6

10 tested gpt-image-2 prompts across portraits, infographics, products, characters, and more, 2026. Quietly useful. https://james-palm.medium.com/chatgpt-images-2-prompts-10-styles-guide-2026-510f76f8141e

#2026 #ai-image-generation #chatgpt #gpt-image-2 #prompts #styles-guide #use-cases

Complete List of Styles & Prompts for ChatGPT Images 2.0

10 tested gpt-image-2 prompts across every real use case portraits, infographics, products, characters, and more (2026 )

Medium

10

Kibotu · May 4

Build a modern LLM from scratch, every line commented, explained like you’re five. No fluff, just solid, practical work. https://github.com/raiyanyahya/how-to-train-your-gpt/tree/master/chapters

#deep-learning #gpt #llm #machine-learning #tutorial

how-to-train-your-gpt/chapters at master · raiyanyahya/how-to-train-your-gpt

Build a modern LLM from scratch. Every line commented. Explained like we are five. - raiyanyahya/how-to-train-your-gpt

GitHub

25

Kibotu · May 3

Your AI doesn't owe the cloud a thing. Thoth runs entirely local: voice, vision, browser automation, knowledge graph, shell, health tracking. Sovereign by design, no data leaves your machine. https://github.com/siddsachar/Thoth

#ai-assistant #ai-tools #automation #browser-automation #knowledge-graph #local-first #personal-ai #voice-assistant