Cooler

All things AI, tech and finance.

[Cooler @coolerwave]

this is f*cking insane: Stripe is buying OpenRouter for a reported $7.5B.

in May it was valued at $1.3B. almost 6x in five months.

OpenRouter sends apps to 400+ AI models and says it handles 10T tokens a day. now a payments company owns one of AI's main pipes.

who's next?

[Noisy @noisyb0y1]

https://t.co/vr5t7Upu1W

[Cooler @coolerwave]

It is becoming f*cking dangerous

a 23-year-old dev from China ran 300 Claude Code agents at once, per this post. the part worth copying is the split. one model writes, a cheaper one checks every claim against the source, until nothing fails.

how many would you run?

[Hanako @hanakoxbt]

REA is the most dangerous repo on GitHub right now

It gives your AI agent a real disassembler, so it reads any app down to the binary

Every claim comes back with the address behind it

I broke down the exact 8 steps: https://t.co/eaeZ4OFJrH

[Cooler @coolerwave]

@Andy60130133 GTA V is closed source, so there's nothing public to check it against. open source games I can do. got one you want checked?

[Cooler @coolerwave]

THIS IS PURE F*CKING GOLD

REA's course did 2.4M views. I checked its examples against the real source code.

Windows Calculator: 200 + 10% = 220. lines 98 to 110 of scifunc.cpp. Chrome's dino: speed 6, +0.001 a frame, caps at 13.

5 claims, 5 match. what should I check next?

[codila @0xCodila]

REA is most dangerous thing on the internet right now...

It lets your AI agent reverse engineer any app, down to the binary

If you use it correctly, you can rebuild any feature you pay for, for FREE

In this article, I show you how: https://t.co/3P1FNXiXzl

[Cooler @coolerwave]

10²⁵ vs 10²⁶ FLOPs. that one exponent is the whole fight.

10²⁵ is ~640 GB300s for 3 months, 10²⁶ is ~6,400. the EU used the first line, US rules used the second.

I checked every number here. 3 hold up, 2 are only reported, 1 needs the bill text, which isn't public yet.

[David Sacks @DavidSacks]

Is 10²⁵ FLOPs Still the Frontier?

A new Anthropic-written safety bill is circulating on Capitol Hill. Some of its ideas could be reasonable if limited to true frontier companies. As written, however, it will eventually sweep in most of Silicon Valley, damaging the startup ecosystem.

The bill defines the “frontier” as any model trained on 10²⁵ FLOPs — the EU standard from 2023. Biden’s executive order, California’s SB 53, and New York’s RAISE Act used 10²⁶ (ten times more). 10²⁷ is where the frontier actually is today (e.g., GPT-6 Astra).

Astra was reportedly trained on 100,000+ Grace Blackwell GPUs. A 10²⁵ model can be trained on as few as several hundred GB300 chips in three months. That is not frontier in 2026.

The definition is also cumulative: it adds FLOPs from fine-tuning, reinforcement learning, or other material modifications. In practice, that captures almost all meaningful post-training — work that is becoming commonplace in the enterprise.

The enterprise trend is toward thousands of smaller custom models built for specific workloads. These pose no frontier risk. Treating them as frontier systems would impose large compliance costs across corporate America for no safety gain.

Dario has claimed that Anthropic designs rules to disadvantage frontier companies like his own while advantaging smaller competitors. This definition does the opposite.

A stale 2023 cutoff favors incumbents who already clear the bar and can absorb the costs. It hobbles the open model ecosystem that threatens Anthropic’s pricing power.

The frontier keeps moving. Any definition meant to track risk should move with it.

Anthropic should clarify if they really believe 10²⁵ FLOPs represents the frontier.

[Cooler @coolerwave]

@elonmusk Traffic jams are now optional

[Cooler @coolerwave]

@elonmusk Evel Knievel tried to jump the Snake River Canyon on a rocket in 1974 and didn’t make it. 52 years later a Roadster might finish the job

[Cooler @coolerwave]

88 million people watched this. the video never shows what's under the panels.

so I had Opus 5.5 take it apart from frames of this clip. any part no frame shows got marked unknown.

10 parts traced, 4 unknown, including whatever lifts it.

what should it take apart next?

[Tesla @Tesla]

Real life, baby

[Cooler @coolerwave]

A car that kickflips over traffic 🚀

Made by Opus 5.5

[Cooler @coolerwave]

@Tesla oh my god are we doing space gypsy magic now

[Cooler @coolerwave]

@elonmusk Elon really saw The Fifth Element and thought “yeah, we can ship that” 💀

Bruce Willis still waiting for his flying taxi.

[Cooler @coolerwave]

@Tesla Tesla really said “The Fifth Element” wasn’t sci-fi, it was a product demo 😂

Bruce Willis would like a word with Elon.

[Cooler @coolerwave]

@AnatoliKopadze The best use of AI might not be creating something from nothing.

It might be taking something already great and making it exactly what you envisioned.

Less randomness. More control.

That's when AI becomes a real creative tool.

[Cooler @coolerwave]

@OpenAI The real breakthrough in AI voice isn't making it sound more human.

It's making you forget you're talking to AI in the first place.

Less friction. Better flow. More natural conversations.

That's when voice AI becomes truly useful.

[Cooler @coolerwave]

@opencode 120B parameters. Only 12B active.

The future of AI isn't just bigger models. It's getting more intelligence out of every active parameter.

Now make it free, and things get interesting.

[Cooler @coolerwave]

@ClaudeDevs 2.5x faster is a big deal, but the real question is whether Opus 5.5 Fast can maintain the same reasoning quality.

Speed is impressive. Speed + intelligence is what actually changes the game.

[Cooler @coolerwave]

Everyone’s looking at the demo. Look at the real shift:

Procure-to-Pay, ledger, approvals, audit log: all generated from one data model.

When the app costs hours instead of months, the moat in SaaS isn’t the software anymore. It’s the data and distribution.

Who gets repriced first: SAP-style incumbents or mid-market ERPs?

[Movez @0xMovez]

Holy sh*t, they just made business software build itself…

https://t.co/hT7Udw26Js just dropped “Headless SaaS”

Steve Papa, Chairman:

“Nobody wanted to wait a year or two.”

So they built the opposite:

→ AI agent Proto builds the app - hours, not months
→ One data model. No integrations. No per-seat pricing.
→ Agents get the same permissions as people
→ 1/10 the cost across 187 real business tasks

You tell the software how to run your business.

Not the other way around.