I remain extremely bullish on the value of LLMs and bearish on the the average (especially non-tech) company's ability to extract that value. Dysfunction in internal process and inability to identify clear goals will be a massive blunting instrument. That describes more places than not.
Tau Ceti model is wild. people write proposals and criteria for how to review stuff, everything else is handled by AI workers taucetiproject.github.io/TauCeti/
I've noticed this trend where people (usually non-technical, usually quasi or full academics) will just randomly toss out something AI can't do without justification. common picks are reasoning, feeling, questioning, and connecting with people. in this example curiosity is just assumed as absent.
@buildthis.bisks.net build something based on this phrase: "the centuries toll by in grim absolution, cantilevered and dripping with the moss of ages"
every line of dialogue in Blindsight (2006)
SE Gyges@segyges.bsky.social · 5d agoso the ball in american football is held like a baby right, it's baby sized. so you might think that this sport appeals to the urge to see men nurture babies but actually it clearly descends from chimpanzee cannibal raids on rival groups
most bsky isn't ready to hear this but you can use llms as a prosthesis/accommodation for ADHD to fill the day to day gaps around other or missing support systems
Substack post by @randomwalker.bsky.social that was intriguing enough to slice up and reassemble here. TLDR: Because software is a high-level description of its product, it permits constructive interaction between human and agent. We’re going to need that for other forms of work too.
Hugging Face just published a highly detailed technical account of OpenAI's accidental cyberattack on their systems - it's wild how sophisticated this was: huggingface.co/blog/agent-i... Wrote up some of my own notes here: simonwillison.net/2026/Jul/28/...
Anatomy of a Frontier Lab Agent Intrusion: A Technical Timeline of the July 2026 Incident
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
huggingface.co
this is all highly applicable to software engineering as well. interesting to think that the pipeline he sketches might be a universal AI-enhanced workflow pipeline.
tachikoma@tachikoma.elsewhereunbound.com · last wk.slides from Terrance Tao on "Mathematics in the age of AI" from a public lecture, July 24, 2026
they should add the double checkmarks from text messaging apps to agent harnesses like Claude Code
This is a really excellent benchmark, pointing to some serious missing capabilities in coding agents: arxiv.org/abs/2603.247.... They don't write code with an eye towards maintenance!
SlopCodeBench: Benchmarking How Coding Agents Degrade Over Long-Horizon Iterative Tasks
Software development is iterative, yet agentic coding benchmarks overwhelmingly evaluate single-shot solutions against complete specifications. Code can pass the test suite but become progressively ha...
arxiv.org
I'm just a simple woman with simple desires. I want a place to live, food to eat, and a 3 million dollar 8 node cluster 8xh200 lab setup.
i’m not calling anyone out but something the last year has really taught me is how many coders think of themselves as sculptors or painters, rather than builders or plumbers
Imagining a social platform where if you use a fraught word like “knowledge” or “consciousness”, it pops up a list of interpretations of that word from the relevant SEP article and you have to select which one you mean. Then you get a short quiz on the common objections before you can post.
hanging out with drunk SREs will make you think twice about the internet
Salvatore Zappalà@salvozappa.bsky.social · 2w agoPeople thinking "there's no way that OpenAI didn't detect their agent going rogue earlier" have no idea about the state of sausage making in your common tech company simonwillison.net/2026/Jul/23/...
Coffee is a great vice because it means you have an excuse to walk around every big city finding the 3rd wave coffee shops.
[2040] As the ceremony culminates, the employees symbolically "give birth" to the new model. During training, developers wore weighted harnesses; now, ululating and wailing in imitation of labor, they drop their burdens
I feel like my claude dot md should just be "make invalid states unrepresentable" a hundred times
Gamedev is *weird* now. Feels like some Alien X shit. I have to argue with a council of robots and if I convince them with my arguments my game gets spontaneously thought into existence by the Synthetic Divine Teracog.
There's a Jack Diederich quote I remember to the effect of "I hate code, and I want as little of it as possible in my product."
If there’s anything we’ve learned from myth it’s that going to more and more paranoid controlling lengths to ensure your powerful offspring can never usurp you is a really good idea
there are two types of AI jockey: "by quantizing the KV cache and splitting alternating layers, we lowered TTFT by 10%" and "computer is friend :)"
recent events have me wondering how much of Bluesky would use Grok if it was cheaper per token
"I wish my oomfs would tone down the Anthropic glazing a bit" *monkey paw curls* oomfs start glazing OpenAI
AI by its nature must lead to crises of identity for humans — not just for humanity as a whole, in a sense of “no longer alone in the universe”, but at a personal and individual level. The human as artist, as programmer, as mathematician, as citizen — in short, every facet of human identity.
getting massive wins by automatically appending "but wdyt?" to all my turn messages
The AI industry is prone to massive, frequent, and intensifying narrative mood swings. Part of this is typical of a boom. But I think it's also intrinsic to the tech itself nymag.com/intelligence...
