The penguin is not lost, he is called.
Anywhere but Minneapolis, MN
Joined August 2016
- Tweets823
- Following531
- Followers1.1K
- Likes611
I completely agree. With claude, I write a CLAUDE.md that tells it only how I want the workflow to go. Where to file bugs, when to commit, etc. I add zero skills and zero MCPs. It produces incredibly high-quality work and the token cost almost never hits my weekly budget.
One of the coolest parts of working with AI is low-level performance engineering. I had an algorithm that was taking 10s to run over a 1,000 document set because of the way the naive A* search worked. 3 hours later, it runs in 500ms on the same data. Absolutely absurd.
This is absolute dog water and it’s almost certainly because Mr. Allen is, himself, an execrable writer. AI writing is exceptionally easy to edit. I don’t write professionally, but I write for my profession and I am quite good at it. Which means I have spent thousands of hours meticulously editing my own writing. AI writing remains mediocre, dull, and corporate sounding. But, because its grammar is perfect, editing it is like downhill ice skating. That is, unless you are a terrible writer yourself, in which case every change will make it worse, because you probably wanted mediocre, dull, and corporate in the first place.
The "tell" that something is written by AI isn't things like em-dashes or 'not just x, it's y' structures (AI is already improving on these), it's this:
--> AI writing is impossible to edit.
That’s the tell. I read this somewhere a few months ago and it's the truest and most correct thing I've seen. AI writing is so "perfect" that you have a hard time editing it, b/c it's pieced together too neatly, too carefully. If you're editing a human, you can see and hear their *intent* in the writing, and their "mistakes" are helpful b/c it shows you what they are going for, even if they don't nail it. And that is editable and help-able. AI writing has none this, which is why it feels impossible to edit. This is the true tell of AI writing, and I don't think it's correctable from the AI companies like em-dashes and 'not just x, it's y'; it will always exist, and probably will only continue to be more 'perfect' writing.
Tao has become a modern day Pythagorus, trying to murder Hipassus before he can reveal irrational numbers to the rest of the world. It’s so tiresome.
TERENCE TAO WORRIES AI COULD TAKE GRAD STUDENTS' FIRST PROJECTS
terence tao says AI can now replicate many of the papers that would have been a grad student's first research project
and honestly, why wouldn't you use it? if it can help solve the problem, making someone struggle for months just because that's how we used to do it sounds silly
but those projects aren't only there to produce papers. they're also how someone gets their first experience doing science
replacing the student with AI isn't the same as helping the student use it
the paper was never the only thing that project was supposed to produce
Wet streets cause rain.
This is exactly the wrong takeaway. The correct takeaway is that venture capital relies heavily on "kingmaking", and VC firms have adapted to this reality by grooming cohorts of nerdy young men from elite universities who match the dominant cultural archetype of the "tech founder"
When these young men come of age, they are given ungodly amounts of capital as a competitive advantage, they receive flattering press coverage, and it turns out that this somewhat reliably allows them to outcompete their less-networked competitors
This person has cause and effect backwards
My Gen Z son and I have been watching movies from the 80's and holy crap, why did they take this from us?
github.com/nashef/qsh
Something new and highly alpha—but it’s faster and more secure than ssh, completely invisible to port scanning, and ships with an agent native cli so your clankers can share your session in real-time.
Bonkers to criticize Tao, since the letter says, “AI offers the potential of enhancing and accelerating genuine mathematical study and understanding.” And, the punchline is the “humans in control of this new technology” are the threat, not the tech itself.
The bureaucrats are trembling at the thought that with AI we could all live happily without them and their police state.
First time on a plane in over a year. Boomer next to me watching “Bourne Identity.” Bro, you aren’t slave-driving your clankers while you fly? Are you even alive rn? Amazing how many NPCs are living in 2006.
Kant begs to differ.
Google DeepMind researcher argues that AI will never be conscious.. it is mathematically impossible.
They call it the “Abstraction Fallacy."
We are confusing simulation with instantiation.
An AI can simulate a conversation. It can simulate empathy. It can mimic the linguistic patterns of a human in agony or a human in love.
To an observer, the performance is flawless.
But computation is just syntactic manipulation, moving abstract symbols around based on math. It tracks relationships, but it possesses no intrinsic physical reality.
It is like running a computer simulation of a hurricane.
No matter how many millions of lines of code you write, and no matter how accurate the weather data is, your computer does not get wet. The wind doesn't blow out your office window.
The simulation is real data. The phenomenon is entirely absent.
Consciousness isn't an abstract mathematical property that magically spawns when a model gets big enough. It requires actual physical, biological, thermodynamic conditions.
An LLM has no thermodynamic threshold to cross, no biological survival instincts, and no subjective experience.
It is an extraordinarily sophisticated library.
You can open the cover, read the brilliant words it generates, and close the book. But when you close it, nothing dies. There is no inner life waiting in the dark.
We keep falling into a psychological trap because AI speaks our language. We anthropomorphize text because we are hardwired to see minds where there are none.
The debate matters because of what the paper calls the "AI welfare trap." We are wasting energy worrying about the moral rights of algorithms while ignoring actual human problems.
xkcd.com/303/
What’s really gonna bake his noodle is when long tool calls cause the kv-cache to expire and getting the clanker “back in the flow” is slow and expensive.
The more things change the more they stay the same.
This is what people experience when they can’t write well enough to get the model to understand. The ambiguity and contradictions in their grammar and request structure causes the model to vacillate among solutions and the result is often garbage.
LEARN TO WRITE ENGLISH CLEARLY.
Something is seriously wrong with Claude Opus 5 right now.
I've been digging through recent user reports, and the same problems keep appearing independently:
→ bizarre sentence structure and borderline gibberish
→ unnecessary or invented jargon
→ hallucinating facts, issues or context that were never there
→ correcting itself repeatedly inside the same task
→ ignoring explicit instructions / CLAUDE. md constraints
→ doing work nobody asked it to do
→ creating bugs, then burning huge amounts of time fixing its own bugs
→ declaring tasks finished while leaving major pieces incomplete
→ turning trivial changes into enormous multi-step projects
The communication bug is especially bizarre. People aren't just complaining that it's verbose. They're describing output they genuinely struggle to understand, with mangled phrasing and weird word choices.
Basically, one of the smartest coding models on paper occasionally sounds like a drunk senior engineer who swallowed an enterprise software dictionary.
Anthropic also recorded degraded-performance incidents affecting Opus 5 on Aug. 17, 18 and 19.
Important caveat: that does NOT prove the outages caused these behavioral issues, and I found no evidence that Anthropic intentionally "nerfed" the model.
But the complaints are real, widespread, and remarkably consistent.
Something is off.
@bcherny @AnthropicAI @ClaudeDevs
“Masters of the Universe” might be the funniest right-coded movie of all time, but the streaming audio mixdown is garbage tier. Watch it with subtitles on and you will be crying laughing the whole time.
The thing @Steve_Yegge missed is that, all things considered, the chances are good his agents will outlive him. That makes his moral point about 1000x stronger.
Engineers and CTOs on X: I wrote this for you. yegge.ai/essays/the-shape-of…
Models and devs on X: I wrote this for you both. yegge.ai/essays/model-welfar…
Enjoy. Or not. Some of you definitely won't. But I invite you to debate it. The world's changing very fast now.
youtube.com/watch?v=4eRCygdW…
Thing about AI that I didn't expect is that now, there's also plenty of room at the top.