In early March, due to the unexpected war in the ME, I was contemplating my next move. Should I stay or leave? Flights were still taking off from Dubai, albeit at a limited frequency. Enterprising car rental companies were offering package deals to wealthy exp…
the good, the bad, and the very ugly of the framework laptop
The odds of a perfect March Madness bracket are 1 in 9.2 quintillion. Warren Buffett bet a billion dollars on it, and he was never worried about paying out.
Google AI Edge Gallery-app - App Store
Download Google AI Edge Gallery van Google in de App Store. Bekijk schermafbeeldingen, beoordelingen en recensies, gebruikerstips en meer apps zoals Google AI…
Note: English is not my first language. The text below was written by me and corrected by an AI. I then corrected the AI’s corrected text, so there are probably still some wording errors In 2022, I made a New Year’s resolution to switch from Chrome to Firefox,…
Years ago I watched a video where guitar teacher Justin Sandercoe explained a way to get better at guitar. It has changed my playing, and it might change yours, too.
The Codification Treadmill
The person who replaces you probably won't be AI. It'll be someone from the next department over who learned to use it.
Good APIs Age Slowly
Good APIs do not win on first impression. They survive change. This post looks at why stable APIs expose less, assume less, and age better over time.
Three tips on how to be more lucky.
Apr 5, 2026
How I fine-tuned Whisper to recognize my atypical speech, dropping WER from 27% to 6.5% by freezing the decoder and only training the encoder.
LLMs don't have feelings, but they have a dataset based on human interaction. It's probably not a great idea to behave in a way that produced negative results in the dataset.
The Anatomy of SMS Delivery: Routing and Carrier Flow
Most SMS APIs hide routing. This guide explains how SMS delivery really works, from API request to pricing, execution, tracking and carrier outcome
How Claude Code Builds a System Prompt
Context engineering in action.
Comments are the most interesting part of the internet.
Design aggregates around true consistency boundaries, not around all related data, to keep domain logic clear and write paths performant as the system grows.
LLMs have eliminated Bazel's biggest adoption barrier — the steep learning curve. They're fluent in Starlark, can generate BUILD files, and make getting started surprisingly easy.
The browser is eating itself
What is a web browser in a post-HTML and post-document era?
Code freezes
Code freezes can undermine the very thing they are meant to protect.
My university uses prompt injection to catch cheaters
Web developer and cybersecurity researcher
_-_-_-_-_-_-_-
Upgrade Your Workspace with FlexiSpot! Enter my exclusive code 'KAZE50' and save an extra $50 on the C7 Max chair! Shop through Flexispot C7 Max：https://bit.ly/4cYtnPr -us https://bit.ly/4dCNMd4 -ca Patreon: https://www.patreon.com/Kazestuff Streams: https://…
Sad Story Of My Google Workspace account suspension
As the title suggests, I have finally become the victim of Google’s account suspension.
For eight years, I’ve wanted a high-quality set of devtools for working with SQLite. Given how important SQLite is to the industry1, I’ve long been puzzled that no one has invested in building a really good developer experience for it2. A couple of weeks ago, …
Someone at BrowserStack is Leaking Users' Email Address
Like all good nerds, I generate a unique email address for every service I sign up to. This has several advantages - it allows me to see if a message is legitimately from a service, if a service is hacked the hackers can't go credential stuffing, and I instant…
GitHub - JuliusBrussee/caveman: 🪨 why use many token when few token do trick — Claude Code skill that cuts 75% of tokens by talking like caveman
🪨 why use many token when few token do trick — Claude Code skill that cuts 75% of tokens by talking like caveman - JuliusBrussee/caveman
On AI agents, grunt work, and the part of science that isn't replaceable.
The Wallet Unit provides for authentication means which can be bound to multiple identification means, such as the PID, via a public/private key pair, see cryptography.
An Amazon/AWS engineer raised the alarms on Friday over the current Linux 7.0 development kernel leading to the throughput for the PostgreSQL database server being around half that of prior kernel versions
It's not that deep
I have these Sunday evenings where I find myself sitting alone at the kitchen table, thinking about my life and how I got here. Usually, these sessions end with an inspiring idea that makes me want to
Programmers tend to fight about why Object-Oriented Programming (OOP) is good or bad. Among the anti-OOP crowd, I often see junior programmers hate on OOP and “rebroadcast” what they’ve heard experienced programmers say. But when challenged to explain why OOP …
Throughout my time working with complex distributed systems, I have learned that consistently detecting systemic issues is surprisingly difficult and a bit of an art. The canonical way to do this is to set up alerts that notify an engineer when something is wr…
An introduction to Turing machines and computation
Contents Introduction Should you read this? The Turing machine The intuitive approach A mathematical approach Languages Words/Strings Languages from words Distinguishing Turing machines with respect to a language A recogniser A decider Distinguishing languages…
Writing formal proofs for distributed protocols is tedious and usually not worth the effort for real-world systems. In general, you have to come up with an inductive invariant which is already a very difficult task. At least a few separate PhD theses were enti…
IDE or ADE, or money printer
Three failed pipeline architectures, a lesson about backpressure, and the UAT gate that finally made multi-AI vibe coding work. An experience-sharing post about what broke, what survived, and why knowing what you want matters more than the tools you use.
Just because it's work shaped doesn't make it productive
AI Assisted/Agentic programming are pretty common place at this point. The growing sentiment seems to be that if you can't find some sort of benefit in your workflow, it's more of a skill issue than a problem with the tools. Whether you believe this to be true…
blazelight.dev
I do things on the computer.
Last month I wrote my experience with AI-generated coding on this blog(two months as a vibe coder). I got a lot of positive feedback, so I created another chart for my work up till this months and wrote down the thoughts I had this month.
How to Evaluate Claude Skill Output Quality for Prompt-to-SQL Scenarios
Implementing the Agent Skills eval spec for a prompt-to-SQL Skill: same-session grading, domain-knowledge assertions, and the eval runner architecture.
This is the first part of a guest series by Patrick Nadeau, a member of the Research Engineering team based in Canada. I mentioned it in this recent post. Patrick is a great writer and a very experienced developer who gives a personal, nuanced perspective on h…
Why domain specific LLMs won’t exist: an intuition
Why not have different models for different capabilities? I wondered this myself and thought that it was quite a natural way for the industry to progress...
One developer, team-quality code — powered by multi-agent AI
How HDR images can make logos and highlights appear unnaturally bright, and how to create them yourself.