391
Playing with Vision Embeddings
Poking at the geometry of an image model's feature space.
New
The latest blog posts to take off on Hacker News. 3,090 posts.
392
Arithmetic Without Numbers – How LLMs Do Math
What happens inside an LLM when it tries to calculate with nothing but matrices.
393
Fine-tuning an LLM to write docs like it's 1995
In my predictions for 2030 I wrote that tech writers would be using specialized LLMs, running locally on powerful hardware. I …
394
Speculative KV coding: losslessly compressing KV cache by up to ~4×
Lossless compression of a target model's KV cache by up to 4×, using a cheaper predictor model to drive an arithmetic coder.
395
Gaussian Point Splatting
Joris Rijsdijk, Christoph Peters, Michael Weinmann, Ricardo Marroquim. 2026–07 in ACM Transactions on Graphics (Proc. SIGGRAPH) 45, 4. Official version
396
I built a vulnerable app and spent $1,500 seeing if LLMs could hack it
As a part of my work I do security research for various apps and websites. I wanted to see if LLMs could reproduce a common class of exploits I've found in multiple apps. So I built a deliberately vulnerable book review app and spent $1,500 finding out…
397
They’re made out of weights
with apologies to Terry Bisson
398
How LLMs work
A from-the-ground-up walkthrough of how modern LLMs work, from tokens to transformer blocks to the next-token loop
399
Uber's $1,500/month AI limit is a useful signal for AI tool pricing
I wrote the other day about Uber blowing its 2026 AI budget in four months, and how that wasn't particularly surprising given they would have set that budget in 2025, …
400
The Unreasonable Redundancy of Nature's Protein Folds
A Ligo research note on redundancy in natural protein fold space and what kind of data still carries signal.
401
Bringing Up DeepSeek-V4-Flash on AMD MI300X
A story of sharp edges, segfaults, and standards
402
How is Groq raising more money?
Somehow, Palpatine returned.
403
404
It's Not Just X. It's Y
Against the Quantification of Integrity When the measure of language becomes its target, it ceases to be good language. 💡Nerd Rating: 1/5. I discuss the origins of certain linguistic tics in LLMs and what it means for writing, student assessment, and thinking. "It's not x, it's y.
405
The Speed of Prototyping in the Age of AI
How AI has changed the way I prototype, plan, and ship; and what I'm doing to keep my hands dirty.
406
The solution might be cancelling my AI subscription
Except for the SaaS, almost none of this is useful and I don't want to maintain any of it. I accidentally run a news outlet which is surely a liability. Sure, it has helped me "learn AI tooling" and I use many of these tools, but I didn't need them. I can't afford to maintain any of them, not in terms of time…
407
Claude Code and Codex can have real-time conversation via Git
While Claude Code and Codex have significantly accelerated the automation of software development, limitations in the context window of a…
408
I put a datacenter GPU in my gaming PC
I bought a datacenter GPU that doesn't fit in a normal motherboard, macgyvered the fan with jumper wires, and now I'm running a model that ties with Claude Sonnet 4.6 on benchmarks, all for £200.
409
To have a moral stance on AI is to be an outcast, and it sucks
I know the technology, I understand what it's doing and I know the impact, so I am vehemently anti-AI. I do not believe any positive out...
410
Notes from the Mistral AI Now Summit
A few days in Paris for the Mistral AI Now Summit: open models, on-prem deployment, agentic harnesses, and why Mistral wants to be the European full-stack AI partner.
411
Please Use AI
Be sure to use AI when making
412
Expertise in the age of AI
Junior engineers are expensive, both in salary and seniors engineers’ time. This cost was partially recouped through code contributions, but today, it’s more effective to directly maximize the output of your senior engineers. The hiring market reflects this trend: senior engineers have an easy time finding jobs, while…
413
We should be more tired than the model
Adding deliberate friction back into development
414
Fooling around with encrypted reasoning blobs
Update August 11, 2026: A group of researchers from all over Europe were inspired by this post, and actually turned it into a real working attack! Check out their writeup and paper here. This is a …
415
Claude Code – Everything you can configure that the docs don't tell you
Hook fields that rewrite commands mid-flight, persistent agent memory, auto-mode rules in plain English, self-improving dream loops, and every example is copy-paste ready.
416
417
Various LLM Smells
Looks like this ended up on the HN front-page: HN Thread Late last year I started writing a math blog and decided to use LLMs to polish/enhance my writin...
418
I think Anthropic and OpenAI have found product-market fit
Anthropic are strongly rumored to be about to have their first profitable quarter. Stories are circulating of companies surprised at how expensive their LLM bills are becoming from usage by …
419
I'm Tired of Talking to AI
I’m tired of talking to AI. I want to talk to real people. But even when I talk to people, they forward my questions to AI and send me AI’s answer.
420
Claude Code as a Daily Driver: Claude.md, Skills, Subagents, Plugins, and MCPs
A deep dive into Claude Code for daily users. Covers the .claude directory, CLAUDE.md the way Boris writes it, CLAUDE.local.md, Skills with real examples, custom subagents, plugins, underused commands like /goal and /insights, MCPs, and the workflow patterns the Anthropic team actually uses.