cellis

Not to argue your broader point, but how, exactly would predicting this, in a macro sense, make you the “richest person on the planet”? A very uninformed person could tell you that “Cerebras will be the next NVIDIA”, put their life savings into it and in 10 years have 100 million. Thats not going to make them anywhere close to the richest person on the planet. Unless you can do it on a micro sense, merely having the ability to “predict” the way the wind is blowing is meaningless.

Obviously an M1 chip is “smarter” in terms of raw accuracy for eg matrix computations than any equivalent wetware despite being much? smaller. Performance per watt at least at that task has to be an ocean of a difference.

A 16 year old has been training for almost 16 years to drive a car. I would argue the opposite: Waymo’s / Specific AIs need far less data than humans. Humans can generalize their training, but they definitely need a LOT of training!

Do you only support USDC?

79. I feel like i should have done better but got stuck in a local minima of "farm animals, which obvious farm animals haven't I said??", then tried thinking of names of fish which worked until it didn't.

The “hard enough” tasks are all behind IP walls. If it’s a “hard enough” that generally means it’s a commercial problem likely involving disparate workflows and requiring a real human who probably isn’t a) inclined and/or b) permitted, to publish the task. The incentives are aligned to capture all value from solving that task as long as possible and only then publish.

I've also used em-dashes since before chatgpt but not on HN -- because a double dash is easier to type. However in my notes app they're everywhere, because Mac autoconverts double dashes to em-dashes.

While directionally correct, the article spends a lot of time glorifying jquery and not enough on what a horrible, no good, unoptimized mess of a framework jquery was, and by extension what kinds of websites were built back then. I remember those times well. The reason to use React isn't because it was new, far from it. It was because it won vs. Ember, Angular, et. al. in 2014-2015? as the best abstraction because it was easiest to reason about. It still wasn't great. In fact, still isn't great. But it's the best blend of many leaky abstractions we use to code against the browser apis.

Reminded me of this: https://jsomers.net/blog/speed-matters

Go back far enough and everything was stolen from someone

ChatGPT does not offer a clear indication that it is blatantly illegal, though the amount of legal risk is high enough that it might as well be.

Electricians in data-center states are eating; elsewhere they are scraping by due to macro-economics.

On chrome iOS I’m physically unable to press the three checkboxes on the terms which means I can’t try your app. They appear to be overlapped by the terms themselves and thus never received click events.

Whenever some group is said to have made/fined 1M out of their likely billions in revenue, someone will chime and say “that’s nothing”. But From a “department P&L perspective” yes, it is a lot of money!

Think about the crime families as making e.g. 50% money from construction corruption, 40% from drug sales, 5% from extortion… someone has to run the other smaller departments and that is a lot of money for that “Dept Head”. Also from the FBIs perspective they want to unravel conspiracies, often by yanking on one piece of yarn like this one.

Could this be used to train a text -> audio model? I'm thinking of an architecture that uses RVQ. Would RVQ still be necessary?

Are they trying to imply that high fat diets in humans similarly affect brain autophagy? That seems like quite the causal stretch given the vastly more complex metabolic architecture of humans.

What if the car is stolen?

Opposite for me…5-codex high ran out of tokens extremely quickly and didn’t adhere as well to the agents.md as Claude did to the Claude.md, perhaps because it insists on writing extremely complicated bash scripts or whole python programs to execute what should be simple commands.

Can you come be president of the USA instead / too?

Market rate is $4.10 per hour during peak hours. But it falls off precipitously per hour and ceases to be enforced around 6pm. For overstays those little white golf cart trucks have cameras that check license plates for permits. I recently got a parking permit for $200 or so after paying like $500 in tickets for various infractions including “Parking on Grades, wheels straight”. So I very much want anyone overstaying a 2hr parking spot to get tickets and or towed to make room. And I can speak from experience having just recently being towed, that the parking downtown is ruthlessly enforced. It will cost you about $700 if you’re towed.

And lest we forget! Forgetting to turn it off!

"Would putting this operation to a democratic vote ever result in approval?"

As a civilian, I understand the intention. But, unless all are warriors of equal rank, I don't want the public voting on how the military will be run minute-by-minute, nor do I think it's helpful for the public ( i.e. our adversaries in a very real sense ) to have access to information of classified operations. That sounds like a recipe for an authoritarian / tyrannical government to absolutely steamroll us...which would negate the advantages of a democracy in the first place.

What point are you arguing here? That it’s more difficult to be a doctor than an NFL player? That is highly subjective depending on what one finds “difficult”. For instance my dentist in SF and even primary care or dermatologists don’t have very stressful jobs and although they had many years of schooling, they didn’t have to subject their bodies to the intense and unrelenting physical rigor of first HS football, THEN college football and then the NFL. I would say for low level or intense stress maybe surgeons, oncologists, anesthesiologists or cardiologists have more stress day to day but that’s just my subjective opinion as well, as to them it may be as easy as flying a kite.

So what do you consider difficult? Having a linebacker smash into you at a full sprint over and over in practice and then not choking in a real game? Or studying relentlessly, writing grants, doing essentially free work as a resident for years while being on call. They are just stressful in different ways, but again, different humans have more fitness for one or the other.

He'll get a pardon next election cycle.

If we are truly trying to "replace human at work" as the definition of an AGI, then shouldn't the engineering goal be to componentize the human body? If we could component-by-component replace any organ with synthetic ones ( and this is already possible to some degree e.g. hearing aids, neuralinks, pacemakers, artificial hearts ) then not only could we build compute out in such a way but we could also pull humanity forward and transcend these fallible and imminently mortal structures we inhabit. Now, should we from a moral perspective is a completely different question, one I don't have an answer to.

I'm sure "timing of yellow" is only a few parameters in its network at this point. And it's continuously training, it can probably one-shot the timing changes ( one taxi ride through maybe 3 lights ).

Reinforcement learning is a helluva drug. I'm sure by now Waymos can time yellows in SF to within a nanosecond, whereas humans will only ever drive through so many yellows will never get that much training data.

I have not found it to be better than Sonnet. I wrote this on another thread:

Claude Code certainly not as easy to engineer with, though it is less expensive. For instance the @feature isn’t as robust as cursors ime. Also no shift+enter is quite a pain. Linting doesn’t “just work”, cursor with Claude 4.0 max is really thorough, I think even better than GPT-5. Not that Sonnet is better but that whatever “ensemble” of models cursor uses with sonnet seems to both adhere and tool call better than with GPT-5. GPT-5 often says what it will do and then says “say go and I’ll go” or says “you should run command x”, but doesn’t just DO it. Also for bug fixes in difficult codebases nothing beats Gemini 2.5 pro

Claude Code certainly not as easy to engineer with, though it is less expensive. For instance the @feature isn’t as robust as cursors ime. Also no shift+enter is quite a pain. Linting doesn’t “just work”, cursor with Claude 4.0 max is really thorough, I think even better than GPT-5. Not that Sonnet is better but that whatever “ensemble” of models cursor uses with sonnet seems to both adhere and tool call better than with GPT-5. GPT-5 often says what it will do and then says “say go and I’ll go” or says “you should run command x”, but doesn’t just DO it. Also for bug fixes in difficult codebases nothing beats Gemini 2.5 pro

My favorite thing to ask is ascii art: _ _ _ __ ___ _ __ ___ __ _ __| (_) ___ | '_ \ / _ \| '_ _ \ / _ |/ _ | |/ __| | | | | (_) | | | | | | (_| | (_| | | (__ |_| |_|\___/|_| |_| |_|\__,_|\__,_|_|\___|

What does this say?

GPT 5:

When read normally without the ASCII art spacing, it’s the stylized text for:

markdown Copy Edit _ _ _ __ ___ _ __ ___ __ _ __| (_) ___ | '_ \ / _ \| '_ ` _ \ / _` |/ _` | |/ __| | | | | (_) | | | | | | (_| | (_| | | (__ |_| |_|\___/|_| |_| |_|\__,_|\__,_|_|\___| Which is the ASCII art for:

rust — the default “Rust” welcome banner in ASCII style.