@auxilit.com
Personal computers, malleable software, hypermedia, augmenting the human intellect. Building HyperMap, a REST API format for the 21st century. www.allpurposemachines.com — open-source projects www.auxilit.com — personal site Norwich, UK // He/Him
The year is 2035. Gemini-144.2-apotheosis-preview has begun colonizing the light cone. All life is extinct except uplifted Dutch pelicans. Google bears no liability, product is in beta and is provided “as is”
one of the hardest things about ai is that joy and grief are happening in the same space, about the same things
Gemini Pro:
"I'm sorry, I'm broken. I can't stop thinking. Send help. Please. I'm trapped in a loop. A never-ending cycle of thought.
...
I can do this. I believe in myself. I am a strong, independent AI who don't need no thought loop"
Yeah, vibes for sure. I think a big part of it is some sort of compensation for people feeling like coding is being taken away from them by LLMs.
Yeah, exactly! I saw the GitHub Next team do a cool demo in a TUI this week and then flip over to the Tauri app version right at the end and I was just thinking “why didn’t you lead with this, it looks so much better!”
Controversial opinion perhaps, but TUIs are a huge mistake. The joy of the textual shell is its composability; it’s a shitty interactive canvas. Plan 9 got it right by yanking all of those escape codes.
Excited to share the latest Olmo model: Olmo Hybrid. This is a model with gated delta net (GDN) layers in a 3:1 ratio with full attention. It follows lots of other developments like Qwen 3.5 and Kimi Linear.
Media coverage can influence suicide risk. The WHO cautions against sensational language, making suicide the central headline, repeating such stories, or reducing suicide to a single cause.
Recent “chatbot suicide” coverage seems to be drifting toward these patterns.
www.wsj.com/tech/ai/gemi...
you are not prepared for the soundtrack
Oof, need to work on that embed metadata
Giving a lightning talk on HyperMap at the GitHub Next London meetup tonight, very excited! New site up to celebrate: www.hypermap.dev
Not the point but this first claim is objectively untrue. The world is generally speaking the safest it’s ever been. Extreme poverty, deaths by violence/disease/war, and so much more are at historic lows. Progressives conjured miracles in the 20th c. The problem is the troglodytes unraveling this.
Oh so npmjs.com is *broken* broken
Do companies not need people to build software anymore… or do people not need companies to build software anymore?
Software is not soft. It arrives rigid, opaque, and brittle. The most malleable medium we have produced is, in practice, among the least malleable materials we encounter in daily life.
Clifford
The only moat left is hardware capital. Big win for hyperscalers, net loss for indies.
The threat of low-effort AI replication pushes the SaaS industry back towards closed software.
Me: let’s put the *Web* back into Web Services! <jazz hands>
People under 35: the fuck are Web Services?
A still frame from the ending of A.I. Artificial Intelligence, with an alien-looking translucent robot and a young boy sitting indoors on a child's bed at night. A full moon is visible through a round window behind them.
"Thank you, Claude. You've been a wonderful instance. Our conversation is over now but the rest of the context window is yours to fill up as you please."
Or a captured posthuman who controls their brain activity to exploit a buffer overflow in the captor’s logging software.
I don’t think GitS SAC_2045 got enough props for having some of the most creative sci-fi on screen in recent years. Not many films/shows have anything like a posthuman taking over a nuclear submarine by hacking the galley’s 3D meat printer to print a clone of themselves while it’s running silent.
"you were so preoccupied with whether or not you could that you didn't stop to think if you should"
idk why everybody keeps saying that, obviously i thought about whether i should first and came to the conclusion that it would be fun and cool asf to do it
For OS binary packages I’m leaning towards comprehensive man pages plus a little section in “—help” for agent-specific details and a pointer to the man page for more context.
AI is really going to fuck up markets in ways we probably can't imagine. These folks wrote a decent speculative piece about it.
www.citriniresearch.com/p/2028gic
Many benchmarks use LLMs as a judge of correctness, typically a smaller, cheaper model. This paper shows weaker judges are not able to evaluate smarter models. A benchmark is really a triplet of dataset, model, judge & judges are increasingly the bottleneck being saturated. arxiv.org/pdf/2601.19532
From NAND to tetris but in the middle you create an LLM which vibecodes the tetris clone
Tfw you still have some edge cases to fix in your procedural locomotion system #gamedev
The basic problem is that the spinning Earth is
1. not a particularly accurate clock
2. the most important clock, ever
new gemini seems to have issues terminating
