Christopher

[Tyler @rezoundous]

I refuse to believe the Astra we have now is the Astra that was supposed to be AGI

[Yuntian Deng @yuntiandeng]

Reading this makes me realize that buying AI inference means placing a lot of trust in the company selling it.

Who at Anthropic can read my conversations, under what conditions, and who oversees those decisions? What determines whether private customer activity becomes material for a public report?

Preventing abuse matters, but I'm uncomfortable with the same company selling the service, judging (and lecturing) its customers, and deciding what to disclose about them.

As a paying customer, I get a bot when I need support. Meanwhile, the company publishes detailed reports about investigating how customers use its product. It feels like a restaurant where I can't get any waiter's attention, but management is very interested in observing how I eat.

[Jun Song @jun_song]

Anthropic complaining about distillation is absurd.

They literally scraped the entire web and trained on humanity's rarest books just to burn them after.

They're the ones who illegally distilled all of human knowledge in the first place.

[Token Gremlin @TokenGremlin]

I don’t know, but it feels like AI models from closed labs get released and then catch some kind of disease a few days later and suddenly become dumber.

[Rajath Gowda @buildwithrajath]

🚨 I’m starting to think OpenAI has figured out the ultimate AI marketing playbook.

Look at the pattern. They follow 3 moves:

1. Make users feel the pain.

- Codex limits get tight.
- Users complain.
- Usage burns faster.

Then suddenly… reset after reset after reset. 😭

Users suddenly go from:

“OpenAI, fix this”
to
“Reset Company strikes again”

Even me, to be honest. 😂🔥

Problem → reset → happiness. 👀

2. Turn every model launch into an EVENT.

- AGI claims.
- Huge benchmark numbers.
- “We may have solved a 90-year-old mathematical problem.”

Suddenly everyone is talking about OpenAI again.

3. Now comes the FOMO.

Tibo says Astra demand is “really unprecedented” and they might have to pause new Pro subscriptions.

A $100–$200/month subscription suddenly feels like something you need to grab before the door closes. 💀

Maybe all of this is completely organic.

Maybe none of it is intentional.

But look at the effect:

Pain → relief.
Hype → attention.
Scarcity → FOMO.
FOMO → upgrades.

OpenAI isn't just building powerful AI.

They're getting VERY good at marketing it and make users feel happy.

OpenAI + product + hype + FOMO is a terrifyingly powerful combination. 👀🔥

[Christopher @weofodthegn]

LOLLLL. Hasan’s reaction is amazing 😂😂

[ostonox @ostonox]

They put Hasan's meat on the RNC Jumbotron

[Token Gremlin @TokenGremlin]

People don’t spam replies, open tickets, and dump screenshots of broken products because they enjoy doing it. They do it because they’re desperate for help.

OpenAI is trying to become infrastructure for modern civilization while serving roughly a billion people every week. At that scale, the answer should be better tooling: use your own AI to surface the problems thousands of users are reporting repeatedly, then give the community human attention where possible.

Posting “spamming replies is not a good way to get help” while representing the company just comes off as dismissive. It shifts attention from the broken product to the frustrated user asking for help.

[Tim Jayas @TimJayas]

I rather spend $49 on a PS5 game than spend $1000+ on Astra/Fable5.1 just to build a 3D game which I can play on localhost

right now AI is only good for making quick game demos and not full games like everyone claims

[Andrew @dremnik]

the general consensus seems to be that coding is basically solved.

yet remarkably it doesn’t seem to have produced more than incremental improvements to most of the software we use.

coding is solved, they say, yet we are still using google drive and notion. coding is solved, they say, yet we can think of no better way to coordinate tasks than linear / jira. coding is solved, they say, yet we are still using gmail and outlook. and we still just message each other about everything in slack.

where is the new generation of tools? why is nothing changing?

why are these companies, supposedly in possession of this technology which has ‘solved coding’, hardly producing anything better than what they were creating before?

the great paradox of the moment.

[Peter Girnus 🦅 @gothburz]

I am the payments logistics officer at the United States Treasury Department who has personally touched every version of the free money this country has been promised since 2020, and I am the only person left in this building who remembers how the first one actually went, because the first one is the only one that was ever real.

Spring 2020. Economic Impact Payments, $1,200 a person, funded, appropriated, moving. Then Mnuchin's office ordered the president's name onto the memo line. Not a rubber stamp. A printed signature, the first time a president's name had ever gone on an IRS payment. Commissioner Rettig wasn't told until the request was already in motion. Our engineers had to reprogram the check-printing software and run it through testing before a single check could clear, because the system had never been built to hold a name that wasn't the Treasurer's. 70 million people waited an extra few days for their own emergency relief, during a pandemic, so a signature could finish QA. I sat in the room where someone asked if we could just skip the testing and was told no, we are not shipping the president's name with a rendering bug in it. The actual money moved slower so the branding could move faster. Every version since has been the branding moving faster and the money not moving at all.

Version one was the DOGE Dividend. It didn't come from this building. It came from a 30-year-old hedge fund manager named James Fishback, CEO of a firm called Azoria, who has said publicly that the number arrived to him in a dream. He posted it to Musk in February of 2025: take 20% of whatever DOGE saves and cut it into checks, math it out to roughly $5,000 a household if the savings hit $2 trillion. Musk replied that he'd check with the president. The president checked. Within days it was a line in a press briefing, no legislation, no appropriation, just a number that had been asleep in one man's head 12 hours earlier now sitting on the Resolute desk. By May, DOGE's own public savings tracker read $175 billion, not $2 trillion, which works out to $1,086.96 per taxpayer, not $5,000. Nobody issued a correction. The number just stopped being said out loud. And when Fishback later described himself as having advised DOGE, a former senior DOGE adviser went on the record to say, quote, you didn't advise DOGE. So the origin story of the $5,000 figure still being used today, 18 months and 2 rebrands later, is a dream, a tweet, and a denial. I have that DOGE savings tracker bookmarked. I check it the way other people check their portfolio.

Version two was the Tariff Dividend, announced that November. $2,000 a person this time, lower number, new funding source: tariff revenue instead of imaginary government savings. This one at least had a real number behind it: $195 billion collected in tariffs for the fiscal year. Economists ran the arithmetic before we did. Cato called it fiscal fantasy in print. A deputy chief economist at Oxford Economics warned that stacking a stimulus check on top of the existing tax cuts could overheat the economy. The math didn't work even before the legal problem showed up, and then the legal problem showed up: the Supreme Court struck down the IEEPA tariffs in February, the ones the dividend was supposed to be funded from, and by May we were issuing refunds to importers faster than customs was collecting anything. I watched the ledger cross from positive to negative in the same spreadsheet tab we'd built to track the dividend. We didn't rename the tab. We just stopped opening it.

Version three is the Trump Dividend, the one I'm actually being asked to price this week. Same number as version one, $5,000, announced Tuesday night from the stage of the midterm convention. Quote: if the Republicans win the House of Representatives and the United States Senate, both of them, because of our tremendous economic success, I will issue a dividend to every adult citizen in the United States for $5,000. One condition attached, and it isn't a funding mechanism, it's a spending restriction: the money has to stay in the country, no Canada, no China, no Germany, his words. So the government has now, for the first time in three attempts, specified where the imaginary money is allowed to be imaginarily spent. Commentators started using the word bribe before the applause even finished. There's a real federal statute for this: 18 U.S.C. 597, expenditures to influence voting, a fine or up to 2 years if it's willful. Nobody in this building has said the statute's number out loud in a meeting, and I've listened for it, because there's a comfortable legal argument sitting right there in plain sight: you cannot criminally buy a vote with money that was never appropriated, never budgeted, and has no line item anywhere in the federal government. It isn't a bribe. A bribe requires a bribe.

I priced it anyway, because pricing a number has never once required the number to exist. 270 million adults, $5,000 each, $1.35 trillion, against a national debt that closed this week at $40.1 trillion. It costs 7 times what version two's tariffs brought in at their peak. I put the number in the deck. I did not put in a funding source, because there's a blank cell where that used to go, and I've started to think of that blank cell as the actual through-line of my career here.

Four checks. One real. One dreamed by a guy who isn't even employed by the government he says he advised. One that went negative by court order. One that costs a Senate seat instead of a tax dollar. And I'm still on the invite list for the pricing meeting, because I'm the only one left who was in the room in 2020 when we tested a signature before we tested the money.

[The.Ink, from Anand Giridharadas @AnandWrites]

The same cluster of people constantly telling us we may soon live forever but also all be dead, we’re on the cusp of untold riches but also will have no jobs, the best is yet to come but also this is your last chance to grab resources, is a form of collective psychological abuse.

[Quantіan @quantian1]

Oh ok, well if Sam Altman says he was intimately involved with the entire project and personally vouches that they acted with integrity then that settles things. Case closed.

[Christopher @weofodthegn]

@RpsAgainstTrump Gayyyyyyy

[Christopher @weofodthegn]

@Gloria_in_x_deo @strangestloop lol. I feel like a Kiwi "thuhstle" would be even more fun

[Christopher @weofodthegn]

@ShutItAIIDown @strangestloop Mmmmm, I like "alluvial plain" even better!

[Christopher @weofodthegn]

@strangestloop "Cellar door"

[Lukas (computer)🛡️ @SCHIZO_FREQ]

The AIs solved navier stokes. Or maybe a person solved it with an AI, and then another AI stole it. Or maybe nobody solved it and everyone is larping. It’s literally not possible for you to know. Probably you don’t even know what navier stokes is. I don’t either

[Emu Brigadier @emubrigadier]

No, Sam.

Removing a co-author over their affiliation, or for any other reason unrelated to their intellectual contribution, is academic fraud.

Pressuring someone to do so is academic misconduct.

You have just admitted to it. I hope that's clear.

[Benjamin De Kraker @BenjaminDEKR]

In the last week, most viral news about OpenAI has been

A. models making Blender renders

B. models (maybe) solving a math problem that even mathematicians regard as a curiousity

Do you see the problem?

On both ends of the spectrum, it's performative, not practical.

[Ryan Orhan @rynorhn]

this guy fucking called this months ago.

openai asked researchers to come use their models and do research with them.

now, a few months later:

math researchers getting fucking scooped by the lab they trusted with their work.

what a fucking precedent.