Someone built a C compiler with an LLM for $100 and it boots Linux
A solo dev used an LLM as a pair programmer to write a C compiler from scratch. It compiles itself and boots the Linux kernel. Total AI budget: $100.
Read the note 1 min read
From the notebook
llm.
50 notes on this topic.
Seven LLMs running in your browser tab, no server involved
Someone built a lab where you can test tiny language models that run entirely client-side. No API calls, no backend, just WebGPU and a lot of patience.
Read the note 2 min read
Chat templates are why models say 'as an AI'
A new paper shows the system prompt is not what makes models say 'as a language model'. It is the chat template wrapper that turns raw output into self-aware hedging.
Read the note 1 min read
Monospace fonts for LLM tokens, not characters
Someone built a tool that generates fonts where every tokenizer chunk takes the same visual width. Suddenly prompt length is readable.
Read the note 1 min read
Someone built an LLM leaderboard that knows what you can afford
A live benchmark that ranks models by performance per dollar spent, not just raw scores. Turns out the best model depends on your budget.
Read the note 1 min read
Mercury 2.5 LLM clocks 770 tokens per second
A new language model is generating text faster than most tools can render it. The bottleneck just moved from the API to the browser.
Read the note 1 min read
The JEV prompting pattern works on models that never heard of it
A Reddit user discovered that giving a model explicit instructions to behave like JEV (a fictional expert persona) produces cleaner outputs, even when the model has no training data about JEV.
Read the note 1 min read
Someone is archiving deleted LLM models before they vanish
Pirate Face archives open-weight language models that companies try to delete. Turns out model takedowns happen more often than I thought.
Read the note 1 min read
ROCm still fights for territory AMD already won
Someone built a tool to make AMD GPUs run local LLMs without the usual driver nightmares. The benchmarks show Vulkan beating HIP by 20 percent.
Read the note 1 min read
Thomas Ptacek's LLM writing workflow is the opposite of what I expected
The Matasano founder uses Claude to draft, then rewrites everything by hand. No copy-paste. The LLM is a research assistant that never ships.
Read the note 1 min read
Someone built a chatbot that lives inside Emacs and I am jealous
Yayster is an LLM that runs locally in Emacs buffers. No API calls, no cloud, just a model sitting in your editor watching you code.
Read the note 1 min read
Someone fed 68000 assembly to Claude and got a working Godot port
A developer used an LLM to translate 1993 Amiga assembly into modern game engine code. The surprising bit is how much of it actually worked.
Read the note 1 min read
When LLM conversation memory accidentally becomes static analysis
Someone fed an LLM a codebase to remember, then asked questions. The model started catching bugs the author missed. Not by design, by accident.
Read the note 1 min read
Your local LLM is smarter than you think
Running a model on your laptop does not mean it is broken. The context window might be.
Read the note 1 min read
Watermarked LLM outputs look identical to humans
A quiz shows that even technical readers cannot spot which GPT-4 response has a cryptographic watermark baked in. The detection gap is real.
Read the note 1 min read
Claude wrote a printer driver from scratch
An LLM reverse-engineered a 2008 HP laser printer's protocol and generated a working macOS driver. No human debugging required.
Read the note 1 min read
What happens when you train an LLM on fifth-grade material only
Researchers built an LLM using exclusively elementary-level text. The results challenge assumptions about what makes a model fluent.
Read the note 1 min read
ThoughtDAG lets you edit LLM conversation history as a graph
A new interface shows chat threads as directed acyclic graphs where you can rewrite nodes and re-run paths. Fixes the branching problem most chat UIs ignore.
Read the note 1 min read
Someone built a formal verifier for LLM-written CUDA code
A new tool checks if ChatGPT's GPU kernels are actually correct before you run them in production.
Read the note 1 min read
Someone fit a reasoning model into 14MB
Needle2 runs on phones, smartwatches, and Raspberry Pis. The entire model is smaller than a single photo.
Read the note 1 min read
Hobby programmers reject LLMs because the struggle is the point
Why learning communities ban code assistants while enterprises mandate them.
Read the note 1 min read
LLM-generated CVE reports are polluting security databases
Someone is flooding the National Vulnerability Database with fake SQLite vulnerabilities written by language models, and it is breaking actual security work.
Read the note 1 min read
Training LLMs on an 8GB GPU is now a weekend project
Someone just released a minimal codebase for supervised fine-tuning, direct preference optimisation, and group relative policy optimisation on consumer hardware.
Read the note 1 min read
LLM code review found 41 security issues for three thousand dollars
A security firm ran Claude through GlobaLeaks' codebase and found medium-severity bugs for seventy-six dollars per finding. The question is whether a human would have caught the same issues faster.
Read the note 2 min read
The router everyone built then threw away
Manifest deprecated their LLM router after six months. The reason is not what I expected.
Read the note 1 min read
Someone built a honeypot that catches LLM scrapers in the act
A researcher deployed a fake human verification page that only AI crawlers would fall for. The logs are filling up.
Read the note 1 min read
OpenAI published their Codex security playbook
OpenAI open-sourced the internal security guidelines they used when building Codex. Turns out threat modelling an AI code generator is different from threat modelling a database.
Read the note 1 min read
What LLM labs actually hire for
Research engineer roles at the big labs filter for production ML skills first, paper count second.
Read the note 1 min read
Debian might ban AI-generated code contributions
A Debian general resolution proposes banning LLM-generated patches. The reasoning is direct: you cannot verify the training data's licensing.
Read the note 1 min read
Kids treat chatbots like people more than adults do
New research shows children anthropomorphise LLMs at much higher rates than adults, which changes how we should think about guardrails.
Read the note 1 min read
Cisco released a family of small security models that actually run locally
Antares models are 1B to 8B parameters, fine-tuned on security tasks, and Apache 2.0 licensed. This is not another rebranded Llama wrapper.
Read the note 1 min read
Kimi K3 fixed 15 CVEs that other models would not touch
A Chinese LLM patched critical security bugs in a codebase where OpenAI's models and Anthropic's Claude refused to engage. The refusal problem is real.
Read the note 1 min read
LLMs that remember you are creepy and useful in equal measure
Gwern argues personalised LLM assistants could filter spam, draft replies, and catch your mistakes before you send them. The privacy trade-off is obvious.
Read the note 1 min read
The MCP security audit no one asked for but everyone needs
Anthropic's Model Context Protocol promised to standardise how AI agents talk to tools. A new audit shows most implementations ship with auth disabled by default.
Read the note 1 min read
Mesh LLM runs inference across random strangers' laptops
Iroh built a system that splits LLM inference across volunteer nodes. The networking stack handles dropouts mid-inference. Wild.
Read the note 1 min read
LLM burnout is real
The moment you realise you are debugging prompt chains instead of writing code.
Read the note 1 min read
LLM coding assistants have a variance problem
Dan Luu measured the same coding task twenty times with the same prompt. The variance in output quality was higher than the difference between model versions.
Read the note 1 min read
An LLM that turns inline comments into Git diffs
Sidenote lets you comment on a rendered blog post, then an LLM writes the actual markdown diff. No forking, no pull requests.
Read the note 1 min read
Micro-Agent: smaller models ganging up on GPT-4
vLLM's Micro-Agent proves that three coordinated 8B models can outperform a single frontier model on complex reasoning tasks.
Read the note 1 min read
Someone wrapped every major LLM API in 500 lines of Bash
A GitHub repo called Bash4LLM+ does what Python libraries do in thousands of lines, using only shell builtins and curl.
Read the note 1 min read
Rio's open-source LLM was just a rebadged merge
A city government announced a locally trained language model. Turns out it was two existing models stitched together with the weights renamed.
Read the note 1 min read
Data2prompt stuffs entire projects into LLM context windows
A CLI tool that flattens your data science repo into one massive prompt. Smart filtering meets the 200K token era.
Read the note 1 min read
Someone ran a trillion-parameter LLM on consumer hardware with Optane memory
An enthusiast loaded a 1T-parameter model into 768GB of Intel Optane DIMMs and got 4 tokens per second on a single GPU. Slow, but it worked.
Read the note 1 min read
LLM code smells: hedging language and structural filler
A developer catalogues the tell-tale signs of AI-generated code. The patterns are obvious once you see them.
Read the note 1 min read
LLM agents forget your constraints halfway through the code
New research shows agents generating backend code slowly drop requirements like authentication checks. The longer the generation, the worse the decay.
Read the note 1 min read
Someone finally built a Rosetta Stone for LLM pricing
Models.dev is an open-source database that tracks pricing, context windows, and rate limits across every major LLM provider. No more tab-sprawl to compare GPT-4 versus Claude costs.
Read the note 1 min read
DeepSeek-V4-Flash revives the steering vector idea
Steering vectors let you nudge a model's behaviour without retraining. They fell out of favour when newer models stopped responding to them. DeepSeek-V4-Flash brought them back.
Read the note 1 min read
Torrix: observability without the database tax
A new LLM observability tool runs without PostgreSQL or Redis. That is not a feature list, that is an architecture decision.
Read the note 1 min read
AI found a CVE in its own training set and we are pretending that is a win
Mythos discovered a vulnerability that was already documented in the data it was trained on. The industry is calling this autonomous discovery.
Read the note 1 min read
An AI agent found a real vulnerability in curl
Mythos, an autonomous security agent, caught a buffer overflow in curl that human auditors missed. The tooling works.
Read the note 1 min read