Mental Models for LLMs
Mental Models help you and your AI agent make better decisions. I prepared a dense collection of mental models and thinking tools, ready to include into your AGENTS.md
Can a local LLM beat -O3?
During his keynote at ACCU on Sea in June, Andrei Alexandrescu made some predictions. One of those was the integration of LLMs and compilers. Compilers can generate optimization remarks for LLMs to process. And LLMs can propose optimizations that are difficult…
The deconstruction of mathematical reasoning
A personal website containing information about the author, creative writing, technical satire, technical articles on compilers and logic
Writing code by hand is not an option anymore - one more thing
Writing code by hand is something from the past, but not everything is lost
Niklas Gruhn - The beauty of roundabouts
I have a weird soft spot for roundabouts. Maybe because I'm a software engineer and the roundabout strikes me as the simple, elegant solution while the traffic light intersection is the ad-hoc, over-engineered mess.
A Letter from a Machine Learning Engineer
My thoughts about an AI-related letter I received recently
CrofAI is an OpenRouter wrapper
September 13, 2026
![]()
I stress-tested Meta Muse until its agent control plane started timing out
Four spawn experiments against a black-box multi-agent runtime, six database lock timeouts recovered from its durable trace, and a careful look at what that evidence does and does not show.
When Both Mirror Drives Fail the Same Way | Marc Paul
A ZFS mirror lost both NVMe members to the same firmware bug, one night apart, and I fired the second shot myself. What redundancy actually assumes, and what a UPS quietly breaks.
The Great Unwind
As AI makes code cheaper to replace, teams need a reliable way to preserve how their product behaves through each rewrite.
Improving Throughput by Optimising KV Cache Efficiency for Agentic Workloads
Boosting throughput by leveraging characteristics of agentic workloads to increase efficiency of the KV cache.
notes: email
Email is infuriating until you realise it's just paper that can teleport. Your messages are paper, each reply is paper, threads are just piles of paper, with all the same strengths and weaknesses of letters written on paper.
Ruby on Rails Lost Its Superpower
Rails gave me a career. But the users and programmers it was designed for have changed, and other stacks now offer the same head start.
We are all Product Engineers now
Agents are eating the software development lifecycle from the bottom up — code generation already gone, review and ops following fast. What remains is figuring out what people actually want, which can't be automated or scaled. That job is called product engine…
The Cloudflare products I actually use
An honest report on the Cloudflare products I use in real projects, what each one solves, the gotchas I found, and what I do not use yet.
How Is Compression Prediction? | Luca Lombardo
When does lossless compression reduce to prediction, and what must already be fixed before that equivalence applies?
Major updates for Apple’s software platforms are now available
Apple’s latest software updates introduce Siri AI and deliver helpful new Apple Intelligence features.
This year, I vibecoded my taxes - Carlo Capocasa Tech Blog
Words about software, extra heavy on opinions
Pacing != pacing development
My take on AI model pacing as a framework for release checks and the competitive pressure around model releases.
The Racing Will Continue Until The Beatings Commence — Rheisen X
The people closest to frontier AI are asking for ways to slow down. Competitive, legal, political, and economic incentives still point toward acceleration.
Technology is a Net Good
Version 0.90 The Bad During the pandemic I kept my phone in airplane mode, for no particular reason other than I was mostly in a place where there was wifi, and my phone worked with wifi calling as many modern cell providers operate. Why? Two reasons: cell sig…
Who Aligns the Aligners? Brief Legal Thoughts on the “AI Safety” Fights to Come
Dario Amodei, the CEO of Anthropic, has published an essay – We Must Pace The Frontier – in which he writes: I have worked on AI for the last twelve years because I believe it could dramatically ra…
What Is an Agent Harness? Harness Engineering Explained
An agent harness is everything around an AI model that grounds it in reality: tools, context, guardrails, the loop and verification. How to build one.
Interpreting Pangram
What is AI generated text anyways?
Enforcing Acyclic Hierarchies With SQL Assertions
Introduction Referential integrity constraints have been available since version 7 of the Oracle Database. While these constraints ensure that referenced rows exist, they cannot express more complex integrity rules that span multiple rows or tables. Check cons…
HomeGrown Devlog - Another Step Closer to Release
Wishlist HomeGrown on Steam: https://store.steampowered.com/app/1894560/HomeGrown/ Play my previous game "Equilinox": https://store.steampowered.com/app/853550/Equilinox/ Support the channel on Patreon and get access to the game & code for Homegrown, the cit…
Registration without a phone number - Feature Requests - Signal Community
you have legitimate concerns, but you’re very wrong about this. zero-knowledge proofs are not only the technology behind donation badges and backup payments, but also behind groups. they can’t tie you to a specific donation just like they can’t tie you to a sp…
Vals AI
Private, domain-specific benchmarks in legal, tax, and finance.