layman

A Layman, but special.

[Chris @ChrisGPT]

You have to give it to Ilya and SSI

> no leaks for 2 years
> quiet as a mouse
> no updates
> July 27th - “We reached the point where our research is worth scaling”

The most minimalist AI lab to grace the frontier

[Derya Unutmaz, MD @DeryaTR_]

I'm going to download the full Kimi K3 open weights just in case Anthropic succeeds in getting a ban on open source via the USG.

We need a backup against the forces of the dark side! ☺️

[Polymarket @Polymarket]

JUST IN: Dario Amodei warns authoritarian governments could use advanced AI to achieve “permanent military superiority” & impose unprecedented repression.

[Roshan Rai @RoshanKrRaii]

This is hilariously tragic. 😭

So someone made a page named Reservation Hatao Movement

It gained Millions of followers in just a few hours.

Then it posted that Caste system should be abolished.

All casteists started unfollowing the page. 😭

Movement ended even before start 🤣🤣🤣

[Chubby♨️ @kimmonismus]

Renting GPUs and moving to open weight models took a $1.2M monthly AI bill down to roughly $ 100K.

That is the core of Ben Cera's new episode. Four months of frontier model spend while Polsia went from 500 users to 5,000 paying customers, then about two months of configuration work on rented GPUs, and June came in at around $ 100K. He also takes you inside the data center where the racks sit.

The detail I found most useful: the math only worked because most of his users were asking simple things of simple codebases. That workload never needed the most expensive model available.
Episode 3 is worth the watch if you are building anything agentic!

[Naval @naval]

If there were backdoors and biases hidden in open-weight models, you could count on closed-weight labs to find and reveal them.

[David Sacks @DavidSacks]

Anthropic maintains that it is entitled to train for free on all the world’s output, even if the author objects. But if a competitor trains on Anthropic’s output after paying for it, that is IP theft. The hypocrisy is breathtaking.

[Today Years Old @todayyearsold]

This might be the craziest animal fact i’ve ever learned

[Mark Ajzenstadt @mardehaym]

Cursor charged us $1,382.59 in about an hour. One engineer, one loop, one afternoon.

We caught it the same day, because every engineer's AI spend sits on a dashboard we built.

The dashboard showed us something worse than one bad hour. Most of our token spend was going to work that never needed a frontier model. Formatting. Classification. Extraction. Summarizing a diff.

So we put a gateway in front of every call and sorted by how hard the task actually is. Cheap model takes the easy work if it passes the eval. Frontier model gets judgment.

So many "AI startups" are just burning VC money on frontier model calls for tasks a $0.001 open model could handle.

[Comedyculture.in @ComedycultureIn]

Humanity won 💗🙌🏼

[Mustafa @oprydai]

this is how a physicist sees the world

[Brian Roemmele @BrianRoemmele]

Confirmation of Anthropic’s BOOK BURNING for their AI: The Great Forgetting of the Amnesia Generation

In the quiet warehouses and shipping docks of 2024, Anthropic did not merely train an AI.

It staged a quiet erasure of human knowledge.

While the public was still dazzled by Claude’s fluent paragraphs and polished marketing emails, the company’s engineers and logistics teams were busy executing Project Panama—an internal campaign whose own documents announced the ambition to “destructively scan all the books in the world.”

The method was industrial and irreversible: buy the physical volume SOMETIMES THE LAST COPY, slice its SPINE with a hydraulic cutter, feed the pages through high-speed scanners, then pulp what remained.

Millions of books: scientific research, novels, rare technical treatises, obscure monographs, volumes that had already survived wars, fires, and decades of handling—were reduced to data tokens and then to wet mulch or burned.

They literally burn the evidence of this crime against humanity.

The company itself preferred the darkness. “We don’t want it to be known that we are working on this,” the planning memos warned.

A federal judge later blessed the slaughter. Because Anthropic had purchased the copies and then destroyed them, the court held, the act was “transformative” fair use. Ownership plus obliteration equaled permission.

The ruling handed every subsequent AI lab a legal license to treat the printed past as disposable feedstock.

That license is now being exercised at industrial scale. In 2026, companies like ISBNdb openly market pre-2022 books as the last pure reservoir of human language—text uncontaminated by the synthetic sludge that large language models themselves have already poured into the open web.

Orders arrive for tens or hundreds of thousands of titles at a time. Non-disclosure agreements are standard.

The preferred language is “digital preservation.”

The actual outcome is the same mechanical process Anthropic refined: spine cut, pages scanned, original shredded.

Rare and near-unique copies are disappearing into the same pipeline. Booksellers across continents report sudden, strange bulk orders for specialized, low-demand titles that no ordinary collector would ever buy in volume.

Once the pages are gone, they are gone forever.

A website can be re-uploaded. A bestseller can be reprinted.

The last three surviving copies of an eighteenth-century botanical text cannot.

This is not progress. It is the Great Forgetting.

An entire generation of machines is being taught to speak by consuming and then discarding the physical record of human thought.

The companies that profit from the resulting fluency call the process necessary, legal, and even enlightened.

They speak of curated knowledge and domain-specific expertise while the spines crack and the paper is carted away.

Anthropic did not invent the hunger for data, but it demonstrated openly in private documents and then under court seal how far a well-funded laboratory was willing to go to satisfy it, and how readily the law would ratify the destruction.

Just to gain a temporary AI benchmark advantage.

The amnesia is already here.

Future models will generate elegant sentences to beat Al sorts of tests about history, science, and literature.

They will do so on the ashes of the books that once contained those sentences.

The generation that built these systems will leave its children fluent machines and emptier shelves.

That is the real product of Project Panama and its successors: not intelligence, but a carefully engineered cultural void, papered over with synthetic prose and non-disclosure agreements.

We are left to forget to remember. We are The Amnesia Generation with our brains being put into a meat grinder and fed back to us.

[Kumar Salmania 🔱🕉 @om7690001376373]

#SalmanKhan's dressing sense was far ahead of its time. Even after 22 years, people still want to dress like him.

India's biggest fashion icon. 👑

[Polymarket @Polymarket]

JUST IN: Johnson & Johnson agrees to a $5,500,000,000.00 settlement resolving ~69,000 lawsuits alleging its talc products caused ovarian cancer.

[𝐓𝐈𝐆𝐄𝐑🖤 @beiing_tiger78]

Baap to baap rehga 🔥

[Lorwen Harris Nagle, PhD @LORWEN108]

A Stanford neuroscientist warns high cortisol wrecks memory, enlarges your fear center, and make your brain feel broken.

If I wanted to fix it naturally, I'd do these 8 things every day:

1. Walk barefoot on grass for 5–7 minutes.

[Arpan Kumar Banerjee @Arpan_Kr_Azad]

Reservation isn't failing the people, public policies by the current govt are failing our institutions. The push toward privatization undermines equal opportunity at its core.The national caste census is essential it's time to bring the real picture of our society into the light.

[Om Patel @om_patel5]

UPDATE, ITS WORSE THAN THE CHATS

the same thing is happening with shared artifacts. every app, doc, dashboard and tool people published from claude is also sitting indexed and searchable

people have already pulled up internal company dashboards, full project plans with client details, and someones entire clinical trial summary with real medical data in it

but think about what people actually build in artifacts:

> financial models, cap tables, revenue spreadsheets
> payroll and salary breakdowns with employee names attached
> customer lists and crm exports
> internal wikis, roadmaps and unreleased product plans
> legal docs, contracts and agreements being drafted
> personal health and medication trackers
> tax documents and bank statement breakdowns
> and every dashboard someone pasted an api key or env variable into to get it working

when you publish an artifact your only two options are "only me" or "anyone with the link"

nobody reading that thinks it means "and google will list it publicly." its the same as an unlisted youtube video showing up in search results

and its already past the point of just being google. third party sites have started scraping and organizing these into browsable collections, so deleting it from search doesnt mean its gone

if you have ever published an artifact, go check your published artifacts and unpublish anything you dont want public

link and not indexed were always two different promises. people only ever agreed to the first one

[Neha @Being_Nehu12]

LATEST: Megastar #SalmanKhan celebrating director vamshi's birthday on the sets of #SVC63🔥

[Yulun Du @Yulun_Du]

The most beautiful LLM tech report I have ever written and read. So proud of the whole team. :)