Can Transformers Do Everything, and Undo It Too?
Blog
Build a Signboard
Staff Software Engineer, Grafana Labs
Data Point: ChatGPT vs. StackOverflow vs. Reddit in helping a guy with a tricky web development problem | Pasta's Corner
What was supposed to be a simple web development problem to solve turned into an adventure in seeking help and learning some new stuff
Someone At YouTube Needs Glasses: The Prophecy Has Been Fulfilled
In my recent analysis of YouTube’s information density I included the results from an advanced statistical analysis on the number of videos present on the home page, which projected that around May 2026 there would only be one lonely video on the home screen.
Modular Code with Reusable Standalone Modules
Let’s explore how reusable standalone modules help us implement modular software in a practical and convenient way!
aadil pickle
aadil pickle's internet home
Agents Should Be More Opinionated | vtrivedy
The best agent products aren't the most flexible, they're the most opinionated. Learn why agents need fewer knobs, not more, and how to design around model intelligence spikes.
Mobile browsers see telephone numbers everywhere
Just like Excel seeing everything as a date, mobile browsers automatically interpret many numbers as telephone numbers. When detected, mobile browsers replace the text in the HTML with a clickable ...
The Return of the Viva
Back in 2018, I was a teaching assistant at the University of Bristol (UK) on the 1st-year Introduction to Computer Architecture course. Aside from coursework and exams, the final grade included a viva assessment. Three or four students would be assessed by on…
LLMs in predicaments
18.11.2025 — about 5 minutes if you skim
Context privateering: debugging custom instructions like a pirate
Adding "always speak like a pirate" to instructions for an AI tool is a helpful generic trick for debugging whether those instructions are being read correctly.
How to deploy ChatGPT on your own hardware, the hard way
How I stitched together LM Studio, OpenWebUI, Cloudflare, and Google OAuth to run "BrakeChat" entirely on my own hardware.
How to Host a Blog on a Subdirectory instead of a Subdomain with Cloudflare Workers | David Ma
A guide to setting up your blog on a subdirectory for better SEO and user experience.
How do we keep apps maintained on Flathub? (or building a more respectful App Store)
There have been a few discussions about what Flathub should do to push developers to maintain their apps on the latest versions of the published runtimes. But most of those lack important details around how this would actually happen. I will not discuss in thi…
Use 20x less peak RAM with dp_knapsack_sliding_hirschberg, a new activation memory budget solver for PyTorch
TL;DR: If you use a default torch._functorch.config.activation_memory_budget_solver=dp, you can change it to dp_knapsack_sliding_hirschberg and decrease your peak RAM usage during memory planning by a factor of 20. Currently it's only availab…
How I build internal developer tools inside a small team
There are thousands of resources on how to build good software products. When I was developing my first SaaS product, LemonSpeak, I often found myself in the same loop: Should I implement a new feature or improve an existing one? The question seems easy - but…
What They Don't Tell You About Maintaining an Open Source Project
i built kaneo.app - an open source, self-hosted kanban board. turns out shipping code is the easy part. here's what maintaining it actually looks like.
What a CTO should know about tech | deadSimpleTech
Clausewitz (and I don't think Jomini particularly disagrees with him on this point) stresses understanding the capabilities of various units and weapons systems. What this means is that while you don't need to know how a cannon was made, how to ride a horse or…
Extracting Reddit Data with AI
A few days ago, I stumbled upon a Reddit thread packed with API recommendations. The thread was long, detailed, and exactly what I needed—except I didn’t want to manually comb through dozens of comments. I figured this was the perfect job for AI. Spoiler alert…
Reproducing the results of a computational life experiment, Part 1
This is a post about a computational life experiment.
What Actually Makes You Senior
The one skill that separates senior engineers from everyone else isn’t technical. It’s the ability to take ambiguous problems and make them concrete.Retry
The Local-First Comeback You Didn’t See Coming
Linear’s Secret Weapon and The Quiet Rebellion Against Slow and Stale Web Apps
Google Antigravity Exfiltrates Data
An indirect prompt injection in an implementation blog can manipulate Antigravity to invoke a malicious browser subagent in order to steal credentials and sensitive code from a user’s IDE.
Orion 1.0 ✴︎ Browse Beyond
After six years of relentless development, *Orion for MacOS 1.0 is here*.
Making Crash Bandicoot
As one of the co-creators of Crash Bandicoot, I have been (slowly) writing a long series of posts on the making of everyone’s favorite orange marsupial. You can find them all below, so enjoy.…
Evidence suggests early developing human brains are preconfigured with instructions for understanding the world
Assistant Professor of Biomolecular Engineering Tal Sharf's lab used organoids to make fundamental discoveries about human brain development.
GrapheneOS migrates server infrastructure from France amid police intimidation claims
The GrapheneOS project has announced on X that they are ceasing all operations in France, asserting that the country is no longer safe for "open source projects"
The unpowered SSDs in your drawer are slowly losing your data
SSDs aren't ideal for long-term data storage
Cancer
2025 rolls in with the shadow of a pending biopsy. I forget about it for a little while. They make you wait so very long to confirm anything. I doubt that anything could be seriously wrong. What ar…
jott - five_times_faster
by @bwasti
Different Models, Same Slop?
Are Big Lab Models Converging to the Same Slop?
A Lightweight Way to Guard Against Prompt Injection
Simple, composable guards for safer prompts and cleaner code
With Love to KDE: Take a Moment
I've been using KDE Plasma for four and a half years. The community is sweet and the software is stellar, and I see a bright future for it. I want it to be the best it can be! So, I'd like to talk about a small incident that I want KDE to lean away from.
Demystifying Determinism in Durable Execution — Jack Vanlightly
Determinism is a key concept to understand when writing code using durable execution frameworks such as Temporal, Restate, DBOS, and Resonate. If you read the docs you see that some parts of your code must be deterministic while other parts do not have to be.&…
Prosody XMPP server in the home lab
Posted: 2025-11-23
Django spotlight: SimpleLazyObject & Co.
Turning the spotlight to Django's lesser known useful tidbits: SimpleLazyObject, lazy and examples showing you how to defer work until you actually need it.
Result is the most boring sum type
If you don't see the point, you may be looking in the wrong place.
Three Months
I've been off work for about three months now. It's a good occasion to take stock of how that's going. In a word, it's been great. I'm surprised how much lighter my mood is. It feels like a weight
Claude Agent Skills: A First Principles Deep Dive
Technical deep dive into Claude Agent Skills' prompt-based meta-tool architecture. Learn how context injection design, two-message patterns, LLM-based routin...
Joint-Embedding vs Reconstruction: When Should You Use Each?
A theoretical analysis revealing when joint-embedding methods outperform reconstruction-based SSL, and vice versa.
Jacob Personal
Not knowing is part of the path - Dmitry Kudryavtsev
I've noticed an interesting trend recently: everyone wants answers. No matter where you look, people just want to know the answer: to life, to business, to lov...
An entire PS5 now costs less than 64GB of DDR5 memory, even after a discount — simple memory kit jumps to $600 due to DRAM shortage, and it's expected to get worse into 2026
You can spend $50 more to get an entire PS5 Pro.
Technical Deflation — Benjamin Anderson
Let's buy the fridge next month, honey.
How to implement action sequences and cutscenes
The personal website of Elias Daler about programming, technology and videogames
Evolving Narratives - Trimming Circles
The age of personalized software
Sam's Spot - Sam saffron's web log
Typography in the wild
how we feel emotions with words
I wrote an EF Core Provider
EF Core Database Provider for Azure Data Explorer (Kusto). GitHub The project that I work on has a lot of data, in a lot of entities with a lot of fields, all exposed through an OData API. Naturally, that compels us to know exactly what kind of query shape…
'My Body Is My Passport; Verify Me': FIDO Auth via Subdermal JavaCard Platform | Jack's Brain | the blog of Jack Kingsman
The blog of Jack Kingsman, a Site Reliability Engineer and geek living in Seattle.
LLM APIs are a Synchronization Problem
Maybe the LLM message APIs should be rethought as a synchronization problem.
How LLM Inference Works
When you enter a prompt into an LLM, the model converts your text into numbers, processes them, and returns a response one token at a time. In this article, we go through the journey of LLM inference and see how it works.
Introducing advanced tool use on the Claude Developer Platform
Claude can now discover, learn, and execute tools dynamically to enable agents that take action in the real world. Here’s how.