Mike retweeted
Maybe we should turn off the internet for a few days. That’s a lot of high impact exploits.
Mike retweeted
'Just one more tax bro, I promise, just one more tax and everything will be fine'
🚨 NEW: John Healey is considering introducing income tax on parents who pass their business down to their children in the Budget
[@Telegraph]
Mike retweeted
anthropic’s lead engineer btw.
Hi, I'm Felix. I'm an eng lead working on Claude. I love watching people take flight and do things they never thought possible.
We often describe that as "being more ambitious" when using AI, which is tricky advice to follow. I wrote a tutorial:
felixrieseberg.com/how-to-be…
Mike retweeted
Generational level of cope going on right now by software engineers that AI has not affected them, is not affecting them, and will not ever affect them
The only ones who don't want to admit this are developers who still think they are "special" in how they interact with technology and who still don't dare to face reality (that all the skills they studied and got experience for are now basically worthless unless you are that 0.001% dev who works at frontier AI Labs).
Mike retweeted
True
Marc Andreessen reveals why housing, health care, and education keep getting more expensive while everything else gets cheaper
"There's basically three components to the American dream. A house near great job opportunities, great health care for you and your family, and sending your kids to good schools"
"If you disaggregate the economy over the last 25 years, everything that technology touches basically collapses in price. Television sets, computer games, anything with electronics in it. But if you chart housing, education, and health care, the prices of those things are skyrocketing"
"The government restricts supply, zoning laws and university certifications. And then the government responds to the political pressure caused by the rising prices by subsidizing demand. And of course, if you subsidize demand and a fixed supply, you get skyrocketing prices"
Codex 6.1 Sol is struggling again today. Selected model is at capacity. @thsottiaux time to boost capacity! #codex #sol61 #modelatcapacity
Apple should offer a premium/pro AppStore account for a lot more, maybe $999/year that gets you priority reviews for apps and a human contact for managing the AppStore. #appstore #proappstore #appdevelopers @AppStore
Mike retweeted
Day 4/
We have improved steering to be instant, leading to the model now reacting much faster to adjustments you make, allowing you to course-correct direction in realtime and not have the model waste effort.
Also releasing GPT-6.1 Sol ultrafast. The two work very well together.
Mike retweeted
I work in the AI industry. I burn about 1.5 billion tokens a week. Between personal and business I have 8 AI subscriptions. And I still think I am falling behind the accelerating rate of change. The last three months alone have been a quantum jump in capabilities. Nobody outside the industry understands how insane this is going to be. Like a whole other industrial revolution, but compressed into 10 years. Nobody is ready for the societal implications. It's going to be a 10.0 earthquake plus an F5 hurricane all at once, globally.
Mike retweeted
Ben Affleck reveals he writes Python, understands convolutional neural networks, worked extensively with GPUs, and used his celebrity status to get private looks at Google and OpenAI’s video models
“I’ve always been kind of into computers since I was young. Then, when film started to move from analog film to digital, I became more interested in that aspect of it. The visual-effects workflow for many years has included machine learning, so I can write pretty shitty Python scripts and stuff like that.
“With convolutional neural networks, which were the precursors to what the transformer can do, which is much more computation simultaneously, you would do things like look at what’s called a tensor. That’s the numerical translation of a visual image in numbers, like the batch number, the frame number and the red, green and blue values of each pixel in each frame. It’s just that simple. That numeric is called a tensor.
“You’d use a convolutional neural network to identify patterns that reveal what’s called edge detection or feature extraction, which is identifying patterns well enough to know, this is where the window ledge is, so we can more easily take the green-screen image out and replace it with something.
“That was familiar to me early on because, prior to Artists Equity, I had a small visual-effects company. I’ve worked with GPUs a lot too. The visual-effects guys said, ‘Hey, you should see. There are a couple: Google and this other company, OpenAI, are doing really interesting stuff with transformers in video.’
“I’ve learned that I can actually just call up and go, ‘Hey, it’s Ben Affleck. Can I come see what you’re doing?’ Sometimes people say yes, to my astonishment.”
$ANET Andy Bechtolsheim says AI has made optics demand about 10 times bigger in five years and the industry is only at the "very beginning"
"So I guess I don't need to tell you that AI has been driving this incredible increase in demand for high-speed optics, which is probably now 10 times bigger than it used to be five years ago..."
"And what I want to talk to you today is that we're not at the end of this journey, but rather the very beginning. There's easily another order of magnitude increase in bits needed for the next generation kind of data centers."
"So what I want to talk about first is what's happening at the data center level, and how we as an industry have to get together to solve these problems of this very rapid and high demand growth as fast as possible, since AI can't wait."
______
For full set of takeaways to Andy's AI Datacenter and Optics lecture: firesidealpha.substack.com/p…
Mike retweeted
Completely unrelated to the rest of the ships today, but our models did solve some important math problems and the team documented it in the following blogpost. Remarkable new era of scientific progress.
openai.com/index/sharing-ai-…
Mike retweeted
I feel like I’m frying my dopamine system with all this Claude/Codex work.
Constantly monitoring and instructing, anticipating the work it’s going to produce.
Then killing time waiting, consuming way more algo feed content than ever before.
How do y’all tackle this?
Mike retweeted
Over the next 28 days, each day we’ll either ship one thing that is a clear improvement and relevant for most codex/work users or ship a full reset. Let the improvements begin.
Mike retweeted
There were a lot of things that only worked because there's a limit to the rate at which humans can operate. We're about to find out what all of them are, as they break.
Mike retweeted
Aviation solved AI slop in 1986.
Karpathy is telling people to ask their LLM to write in ASD-STE100, a controlled language built for aircraft maintenance manuals. I went down the rabbit hole on where this spec comes from, and the backstory makes the hack even better.
Back in the late 1970s, European airlines were handing English manuals to mechanics around the world who mostly didn't speak English as a first language. One ambiguous sentence in a fuel line procedure can get people killed. So the industry built a version of English where ambiguity is banned at the dictionary level.
The spec allows roughly 900 approved words. Each word gets one meaning and one part of speech. "Close" can only be a verb, because "close the door" and "the close door" can't be allowed to coexist in a repair manual. "Follow" only means to come after. If you want obey, you write obey.
Procedural sentences max out at 20 words. One instruction per sentence. Active voice only. Around 1,200 common words are explicitly banned, each with a mandated replacement. You don't "commence pumping." You start the pump.
English has about 170,000 words in current use. Aviation decided mechanics get 900, and planes became safer to maintain because of it.
Now flip it to LLMs. Models hedge and synonym-cycle because the training data rewards sounding fluent. STE is 40 years of accumulated rules for stripping exactly that out, written by people whose readers would die if a sentence could be read two ways. It might be the strongest anti-slop prompt in existence, and it's a free PDF.
We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:
Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:
Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:
Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:
Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!
In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
Codex Sol 6.1 seems to be grinding to a halt. I get that it’s supposedly as capable as Astra but what’s the point if it’s running at quarter-speed? Overloaded servers or have @OpenAIDevs nerfed the Pro 20x plans even more? @thsottiaux #codex #sol61 #slow #broken