- Jacobian Conjecture Counterexample
- #agent #lm #context
    - Language model harnesses are compositional generalizers
- #agent/idea #math
    - Counterexample Discovery Method
- #math
    - The new counterexample to the Jacobian conjecture
    - A digestion of the Jacobian conjecture counterexample
    - Human mathematicians are being outcounterexampled
- #math #formal
    - Counterexample to the Jacobian conjecture
    - Jacobian conjecture counterexample verification

- #agent #benchmark #perf
    - Benchmarks Are Dead (for us)
- #agent #context
    - Dreaming: Better memory for a more helpful ChatGPT
- #agent #tui #workflow
    - Vim of Coding Agents | Rasyidan A F
- #agent/idea
    - richard sutton on oak at rlc 2025
- #lm #disect
    - A Visual Guide to Attention Variants in Modern LLMs
- #sec #crypto #agent
    - Let’s talk about encrypted reasoning

- #agent
    - gi-dellav/zerostack: Minimalistic coding agent written in Rust, optimized for memory footprint and performance
    - Codex-maxxing - Jason Liu
    - The last six months in LLMs in five minutes
    - Welcome to Learn Harness Engineering | Learn Harness Engineering
- #agent #rust
    - cargo-crap: Finding Untested Complexity in AI-Generated Rust Code
- #agent/idea
    - Don't Outsource the Learning
    - Do I belong in tech anymore?
    - no slop grenade
    - Your Evals Will Break and You Won't See It Coming
- #agent/idea #mac
    - The Mac App Gold Rush in the Age of Vibe Coding
- #copyright
    - OrcaSlicer-bambulab/bambu_agpl.md at main
- #disect
    - Under the Hood: Building a Real-Time Chord Recognizer

- #fediverse
    - Open Source P2P Network for Social Apps
    - rocksky.app/rocksky
    - Discover European products and services
- #forge #idea
    - Before GitHub
- #forge #jj
    - ejoffe/spr: Stacked Pull Requests on GitHub
- #formal #lean
    - Using algebra and LLMs to verify a flight-plan bug fix in Lean
- #gpu
    - alternbits/awesome-cuda-books: A curated list of best cuda programming books
- #gpu #disect
    - Designing a Scientific Calculator from scratch in FPGA
- #idea
    - Don't answer the first question
    - Holiday Smorgasbord
- #lang
    - XS
- #math
    - Doron Zeilberger's 82nd Opinion

- #agent #loop #diagram
    - Dunik's 4 loop ladder for Loopcraft — Four loops, one useful caveat.
- #agent/idea
    - Loop Engineering
    - Everything That Went Wrong With Claude
    - The solution might be cancelling my AI subscription
- #disect #forge
    - On Rendering Diffs
- #lean #math #qc
    - Quantum Algorithms in Lean
- #news #env
    - Earth is now heating up twice as fast as in previous decades
- #openweb
    - Goodbye RSS
- #os
    - Haiku OS runs on M1 Macs now
- #os #fun
    - The Virtual OS Museum
- #rust
    - Scoped Error in Rust
    - You can read Rust already (until it starts being Rust)
- #rust #zig #bun
    - My Thoughts on the Bun Rust Rewrite
- #rust #zig #bun #benchmark #perf
    - Rewriting Bun in Rust
- #rust #zig #disect
    - Bun's unreleased Rust port has 13,365 unsafe blocks. Most can be removed.
- #sec
    - My domain got abused on Github Pages
- #sec #tui
    - gripebomb/ThreatDeck: Terminal-based threat intelligence monitoring and alerting platform
- #tui
    - strace-ui, Bonsai_term, and the TUI renaissance
- #wasm #sandbox
    - A Self-Hostable Wasm Sandbox for JavaScript Workers
- #web
    - uni-stack/uniwind: From the creators of Unistyles: The fastest Tailwind bindings for React Native
    - Native all the way, until you need text
- #web #agent/idea
    - Tailwind: Suffering From Success
- #web #design #disect
    - The footgun of right-to-left decorative characters
- #zig
    - Returning to Zig
- #zig #async
    - Async I/O in Zig 0.16, today

- #agent
    - A few words on DS4
- #agent #disect
    - DeepSeek-V4-Flash means LLM steering is interesting again
- #agent #paper
    - Self-Distillation Enables Continual Learning
- #agent/idea
    - Comment by zmmmmm: I think AI rescue consulting is going to be come a...
    - A quote from Mitchell Hashimoto
    - Not so locked in any more
    - Post by @steipete on X
- #agent/idea #math
    - Terence Tao (@tao@mathstodon.xyz)
- #anti-bot
    - Amazonbot is finally respecting robots.txt
- #forge
    - GitHub is sinking
- #fun
    - Browse Wikipedia on a Windows XP desktop
- #idea
    - Students Don’t Need a Million Different Explanations
- #idea #selfhost
    - We've made the world too complicated

- #idea #zig #rust
    - Zig vs Rust in 2026
- #jj
    - Git Is Not Fine
- #lang
    - Scripty
- #math
    - John Carlos Baez (@johncarlosbaez@mathstodon.xyz)
    - John Carlos Baez (@johncarlosbaez@mathstodon.xyz)
    - Terence Tao (@tao@mathstodon.xyz)
    - Terence Tao (@tao@mathstodon.xyz)
- #math #agent/idea
    - Terence Tao (@tao@mathstodon.xyz)
- #mm
    - Efficient Minute-Scale World Modeling
- #openweb
    - From RSS to Atom
- #rust #fediverse
    - GianIac/numax: The beginning of a new runtime model
- #rust #shader #gpu
    - The cuda-oxide Book
- #rust #zig #agent/idea
    - A quote from Mitchell Hashimoto

- #sec
    - C/C++ Security Checklist
    - C/C++ checklist challenges, solved
    - Maybe you shouldn't install new software for a bit
- #sec #rust
    - Bugs Rust Won't Catch | corrode Rust Consulting
- #selfhost
    - Where to buy a non-Apple, non-Google smartphone
- #tui
    - peternaame-boop/ytm-player: YouTube Music TUI client with vim keybindings, synced lyrics, and cross-platform media keys
    - simonomi/hexapoda: a colorful modal hex editor
    - Epiq — terminal-native issue tracking
- #web
    - A few ways of specifying per-theme colours in only CSS
    - Moving away from Tailwind, and learning to structure my CSS
- #zig
    - Steering Zig Fmt
- #zig #build
    - peppergrayxyz/c-zigbuild: Utilize the Zig Build System for C
- #zig #rust #agent/idea
    - My Thoughts on Bun's Rust Rewrite

- #agent
    - Evaluating Stochasticity in Deep Research Agents
    - Toward Expert Investment Teams:A Multi-Agent LLM System with Fine-Grained Trading Tasks
    - Task-Aware Delegation Cues for LLM Agents
    - LLMGreenRec: LLM-Based Multi-Agent Recommender System for Sustainable E-Commerce
    - COMIC: Agentic Sketch Comedy Generation
    - What is agentic engineering?
    - Are LLMs not getting better?
    - obsessiondb/rudel: Claude Code & Codex Session Analytics
    - This _all_ (waves hands around) sounds like alot of work and expense for somethi
    - > sounds like alot of work and expense for something that is meant to make pr
    - Some insider knowledge: Lilli was, at least a year ago, internal only. VPN acces
    - What a welcome post.  The whole reason I come here is to get thoughtful input fr
    - For quite a while, I like use LLM to refine and fix my grammar issue, but my col
    - Reliable Software in the LLM Era
    - Thoughts on OpenAI acquiring Astral and uv/ruff/ty

- #agent
    - AI should help us produce better code - Agentic Engineering Patterns
    - Hoard things you know how to do - Agentic Engineering Patterns
    - How coding agents work - Agentic Engineering Patterns
    - [AINews] Anthropic accuses DeepSeek, Moonshot, and MiniMax of >16 million "industrial-scale distillation attacks"
    - [LIVE] Anthropic Distillation & How Models Cheat (SWE-Bench Dead) | Nathan Lambert & Sebastian Raschka
    - ⚡️The End of SWE-Bench Verified — Mia Glaese & Olivia Watkins, OpenAI Frontier Evals & Human Data
    - Qwen-AI Slides is really slept on! It generates PowerPoint Presentations in minutes
    - - YouTube
- #agent #lean
    - Readings shared March 9, 2026
- #agent #prod
    - Post by @cursor_ai on X
- #agent #rag #prod
    - Post by @llama_index on X
- #agent #tui
    - Comment by throwaway63467: It’s a small tool shop building a tiny part of the...
- #benchmark #inference
    - Post by @rronak_ on X
- #claude #agent #prod
    - Post by @claudeai on X
- #claude #benchmark #agent
    - Post by @AlibabaGroup on X

- #fediverse
    - POSSE
- #formal #lean
    - Readings shared March 14, 2026
- #gpt #agent #prod
    - Post by @OpenAIDevs on X
- #gpt #benchmark #agent
    - Post by @OpenAI on X
- #gpu
    - SOL-ExecBench: Speed-of-Light Benchmarking for Real-World GPU Kernels Against Hardware Limits
- #idea
    - tone indicators
- #inference #oss #benchmark #gpu
    - Post by @ollama on X
- #jj
    - Interview With Nathan Goldbaum
- #lean
    - TorchLean: Formalizing Neural Networks in Lean
    - Readings shared March 19, 2026
- #openweb
    - SvelteKit i18n and FOWL
    - ChatGPT, Claude, and Gemini Render Markdown in the Browser. I Don't
- #ph
    - Quantum Information Approach to Bosonization of Supersymmetric Yang-Mills Fields
    - Split or compact?
- #prod #benchmark
    - Post by @Kimi_Moonshot on X

- #qwen #benchmark #gpu #lm
    - Qwen 3.5 27b: a testament to the transformer architecture
- #qwen #oss
    - Post by @JustinLin610 on X
- #rust
    - Rust debugging survey 2026
    - Redirect
    - jtolio/ajail: a basic jail for programs you don't completely trust
    - If you want to know how serious to take this, just look at this gem:<p>&gt; Enha
    - This Week in Rust 642  · This Week in Rust
    - The price of accountability: corruption erodes social trust more in democracies than in autocracies
    - OpenUI - The Open Standard for Generative UI
- #rust #agent
    - Implementing a clear room Z80 / ZX Spectrum emulator with Claude Code
    - What is agentic engineering? - Agentic Engineering Patterns
- #rust #zig
    - Against Query Based Compilers
    - Consensus Board Game
- #sec
    - Six mistakes in ERC-4337 smart accounts
- #sec #mm
    - What creative technical outlets of yours have been ruined by generative AI?

- #mac/wm
    - Native Instant Space Switching on MacOS
- #tui
    - Atuin v18.13 – better search, a PTY proxy, and AI for your shell
    - gptop - A TUI GPU monitor with support for both Apple Silicon and NVIDIA GPUs.
    - ralph-tui -  A terminal UI for orchestrating AI coding agents.
- #zig
    - BearVM
    - ANDRVV/SPSCQueue: A bounded single-producer, single-consumer and lock-free concurrent queue written in Zig, C++17
    - SB0LTD/sbcode: Visual Studio Code Fork by SB0 LTD
    - Proposal: better bit pointer syntax and semantics · Issue #24061 · ziglang/zig · GitHub
    - Noel Welsh:  Parametricity, or Comptime is Bonkers
    - Zig, Zen-C, C3, Carbon or CPP-Front ?
    - Hot Take: Zig is actually very ergonomic
    - Questions about the build system and the community
    - Is learning C necessary before learning zig if it's your first low level language?
    - Devlog
- #zig #lang
    - A ball bounce in Splineworks (a 2d vector animation tool)

- #agent
    - MiniMax-M2.7 Announced!
    - The dictionaries are suing OpenAI for "massive" copyright infringement, and say ChatGPT is starving publishers of revenue
    - Harmonic unleashes Aristotle, the world's first formal mathematician agent for free
    - An Australian ML researcher, used ChatGPT+AlphaFold to shrink 75% of his life-threatened dog’s MCT cancerous tumor, developing a personalized mRNA vaccine in just two months - after sequencing his dog’s DNA for $2,000
- #agent #prod #reasoning
    - An Australian ML researcher, used ChatGPT+AlphaFold to shrink 75% of his life-threatened dog’s MCT cancerous tumor, developing a personalized mRNA vaccine in just two months - after sequencing his dog’s DNA for $2,000
- #agent #reasoning #benchmark
    - Harmonic unleashes Aristotle, the world's first formal mathematician agent for free
- #benchmark #prod
    - Cursor’s ‘Composer 2’ model is apparently just Kimi K2.5 with RL fine-tuning. Moonshot AI says they never paid or got permission
- #claude #prod
    - Claude Pro feels amazing, but the limits are a joke compared to ChatGPT and Gemini. Why is it so restrictive?
- #claude #prod #oss
    - I built a Claude skill that writes accurate prompts for any AI tool. To stop burning credits on bad prompts. We just hit 600 stars on GitHub‼️
- #gpt #prod #agent
    - The dictionaries are suing OpenAI for "massive" copyright infringement, and say ChatGPT is starving publishers of revenue
- #local #inference #agent
    - 128gb M5 Max for local agentic ai?
- #qwen #local #benchmark
    - Qwen3.5 is a working dog.

- #agent
    - We built 76K lines of code with Claude Code. Then we benchmarked it. 118 functions were running up to 446x slower than necessary.
    - What’s everyone actually running locally right now?
    - New Qwen3.5-35B-A3B Unsloth Dynamic GGUFs + Benchmarks
    - Qwen 2.5 -> 3 -> 3.5, smallest models. Incredible improvement over the generations.
    - Trump goes on Truth Social rant about Anthropic, orders federal agencies to cease usage of products
    - Post by @AnthropicAI on X
- #mm
    - Google releases Nano banana 2 model

- #agent
    - The Value of Things
    - Modeling Distinct Human Interaction in Web Agents
    - ProofOfConcept
    - A quote from Thariq Shihipar
    - Minions: Stripe’s one-shot, end-to-end coding agents
    - How Taalas "prints" LLM onto a chip?
    - Bitter Lessons in Venture vs Growth: Anthropic vs OpenAI, Noam Shazeer, World Labs, Thinking Machines, Cursor, ASIC Economics — Martin Casado & Sarah Wang of a16z
    - [AINews] Anthropic's Agent Autonomy study
    - The Scientist and the Simulator
    - Claude Sonnet 4.6 One-shotted this surreal Time-Themed website, full prompt + codepen below
    - GLM-5 Technical Report
    - Kitten TTS V0.8 is out: New SOTA Super-tiny TTS Model (Less than 25 MB)
    - Free ASIC Llama 3.1 8B inference at 16,000 tok/s - no, not a joke

- #agent
    - GGML.AI has got acquired by Huggingface
    - Google releases Gemini 3.1 Pro with Benchmarks
    - Post by @joshwoodward on X
- #agent #crdt
    - Context Graphs: Hype or actually Trillion-dollar opportunity?
- #agent #formal #lean
    - MerLean: An Agentic Framework for Autoformalization in Quantum Computation
- #agent #idea #rust
    - The Human Root of Trust
- #agent #lean
    - Claude Sonnet 4.6: clean upgrade of 4.5, mostly better with some caveats
- #agent #sec
    - What Makes a Good LLM Agent for Real-world Penetration Testing?
- #agent #sec #openweb
    - Using threat modeling and prompt injection to audit Comet
- #agent #tui
    - Automatic programming
    - models - A fast CLI and TUI for browsing AI models and coding agents.
- #formal
    - SMTLIB as a Compiler IR I
- #formal #lean
    - Construction-Verification: A Benchmark for Applied Mathematics in Lean 4

- #formal #lean
    - Statistical Learning Theory in Lean 4: Empirical Processes from Scratch
    - Formalizing Gröbner Basis Theory in Lean
    - Pursuit of Truth and Beauty in Lean 4: Formally Verified Theory of Grammars, Optimization, Matroids
- #idea #openweb
    - </> htmx ~ Prefer If Statements To Polymorphism...
- #lean #formal
    - LeanArchitect: Automating Blueprint Generation for Humans and AI
- #lean #py
    - Calling Lean Functions As Python Functions
- #rust
    - One page of async Rust
    - Adventures in Rust borrowing deserialisation
- #rust #agent #jj
    - Interview With Steve Klabnik
- #rust #zig
    - make.ts
    - CI In a Box
    - Programming Aphorisms
    - Justifying text-wrap: pretty

- #rust #zig
    - Diagnostics Factory
    - Wrapping Code Comments
- #tui
    - whosthere - A local area network (LAN) discovery tool with a modern TUI interface.
- #tui #im
    - sparklost/endcord: Feature rich Discord TUI client.
- #untagged
    - A quote from Martin Fowler
    - Typing without having to type
    - LadybirdBrowser/ladybird: Abandon Swift adoption
- #zig
    - zemit
    - sdzx-1/troupe: Multi-role finite state machine
    - Six months of yak shaving a Zig web backend stack
    - - YouTube
- #zig #agent
    - krillclaw/KrillClaw: The world's smallest AI agent runtime. 49KB. Written in Zig. Zero dependencies.
    - nullclaw/nullclaw: Fastest, smallest, and fully autonomous AI assistant infrastructure written in Zig

- #agent
    - Introducing OpenClaw — OpenClaw Blog
    - Email Inbox API for AI Agents
- #agent #sandbox #selfhost
    - Introducing Moltworker: a self-hosted personal AI agent, minus the minis
- #agent #sec
    - Moltbot Security Guide: Stop Hackers & Vulnerabilities
- #agent/idea
    - Testing Agent Skills Systematically with Evals
    - AGENTS.md outperforms skills in our agent evals
- #cg #disect
    - Cutting Up Curved Things
- #cpp
    - cdervis/cppstat: C and C++ Compiler Support Page
- #wasm
    - 10 Years of Wasm: A Retrospective

- #db
    - Stop using MySQL in 2026, it is not true open source
- #disect
    - How the Lobsters front page works
- #disect #fediverse
    - A Social Filesystem
- #font
    - Loretta1982/xenia
- #sandbox
    - What is Lume?
- #wayland #zig
    - river-classic

- #agent
    - I'm a Happy engineer now
- #agent #sandbox #sec
    - finbarr/yolobox: Let your AI go full send. Your home directory stays home.
- #agent/idea
    - choosing learning over autopilot
    - Three Inverse Laws of AI
- #anti-bot
    - Oh noes!
- #cg #tui
    - ASCII Sketch
    - Asciistrator - ASCII Vector Graphics Editor
- #design #wayland
    - [XFCE] My Xfce points

- #kb
    - pwmt.org
- #kb #tui
    - Han8931/gorae
- #lean
    - Harmonic
- #tui
    - Han8931/gorae at terminaltrove
    - NSPC911/rovr at terminaltrove
    - karol-broda/snitch at terminaltrove
- #tui #browser
    - Text-based web browsers
- #wayland
    - Rubenerd: Xfce is great

- #agent #sandbox
    - A field guide to sandboxes for AI
- #agent/idea
    - OthmanAdi/planning-with-files: Claude Code skill implementing Manus-style persistent markdown planning — the workflow pattern behind the $2B acquisition.
    - Why I Stop Prompting and Start Logging: The Design-Log Methodology
- #cg
    - Quantum Tunnel
- #cpp
    - Clang Hardening Cheat Sheet - Ten Years Later
- #cpp #rust #lint
    - The Story of Building a Rust-Style Static Analyzer for C++ with AI
- #editor
    - I switched from VSCode to Zed
- #forge #pkg
    - git-pkgs: explore your dependency history
- #oss #forge #system
    - Home
- #oss #iot
    - Bose is open-sourcing its old smart speakers instead of bricking them
- #rust
    - Stop Forwarding Errors, Start Designing Them | FastLabs
- #sec
    - Everything You Need to Know About Email Encryption in 2026
- #struct
    - tdortman/Cuckoo-GPU: High-Performance GPU Cuckoo Filter
- #web
    - bigskysoftware/htmx: </> htmx - high power tools for HTML