Priit @ Amperly AI productivity

Latest insights on AI to stay ahead of the competition.

Follow for updates on how AI drives growth, efficiency, and productivity. AI²

[Priit @ Amperly AI productivity @amperlycom]

Agent Arena has GPT-5.5 on top, Claude Opus 4.7 close behind. 300k sessions with real tools, bash, search, files. the rankings barely look like the chatbot arena

[Priit @ Amperly AI productivity @amperlycom]

Anthropic is calling for a global AI pause, worried models could soon improve themselves without human help. feels like we keep hitting these moments faster than expected

[Priit @ Amperly AI productivity @amperlycom]

$100 billion in new issuance pulls on every existing position. the allocation desk and the liquidation desk might as well be the same phone

[Priit @ Amperly AI productivity @amperlycom]

the recursive self-improvement loop is closing. anthropic's latest paper says claude wrote most of the code for its successor, and helped write the safety warning too

[Priit @ Amperly AI productivity @amperlycom]

@paulg what makes LLMs different is that the incumbents aren't late to the party. they threw it. the disruption and the disrupted wear the same badge

[Priit @ Amperly AI productivity @amperlycom]

Anthropic says 80% of their code is written by Claude, ai is improving itself every 4 months, and they want a global slowdown

[Priit @ Amperly AI productivity @amperlycom]

@chetaslua google shipping a weaker public model while a stronger one sits internally changes how you read every benchmark and every comparison between labs

[Priit @ Amperly AI productivity @amperlycom]

Anthropic called for an AI pause while its investors valued it at $965B. hard to ask the race to stop when you are sprinting fastest.

[Priit @ Amperly AI productivity @amperlycom]

anthropic, the company most vocal about ai safety, just filed to go public while asking everyone to slow down. the tension is real and worth thinking about.

[Priit @ Amperly AI productivity @amperlycom]

@unusual_whales putting a global pause demand through wsj means it lands on desks that don't read arxiv. the channel picks the audience

[Priit @ Amperly AI productivity @amperlycom]

anthropic says 80% of its code is now written by Claude, and the timeline for recursive self-improvement keeps pulling forward

[Priit @ Amperly AI productivity @amperlycom]

@AIHighlight capturing a solved workflow as a reusable template changes the economics. you aren't paying for the same reasoning twice, and every solution makes the next one cheaper

[Priit @ Amperly AI productivity @amperlycom]

the flat rate AI era lasted about 18 months. GitHub Copilot killed it june 1, switching to per-token pricing. agentic workflows burn too many tokens for subscriptions

[Priit @ Amperly AI productivity @amperlycom]

US models now charging $5 per million tokens while DeepSeek sits at $0.43. the price gap alone might force a policy response before the tech gap does

[Priit @ Amperly AI productivity @amperlycom]

@ZackKorman every lab that takes the lead seems to discover that safety concerns scale with market position. the timing rarely lines up with the technology

[Priit @ Amperly AI productivity @amperlycom]

Codex now has shareable profiles and an iOS simulator built right into the app. the line between ai assistant and operating system keeps getting thinner

[Priit @ Amperly AI productivity @amperlycom]

@tenobrus a decade of editorial judgment about what clean code looks like transfers perfectly to prompting claude. the keyboard changed but the decisions didn't

[Priit @ Amperly AI productivity @amperlycom]

300 agents, 4000 steps, one afternoon. someone built a working saas without writing a single line of code. the architecture matters as much as the model now

[Priit @ Amperly AI productivity @amperlycom]

@Lentils80 a model generating a functional desktop is a strange artifact. the environment we use to prompt the model is becoming something the model can build

[Priit @ Amperly AI productivity @amperlycom]

claude now generates over 80% of anthropic's internal code, ai writing the ai, the recursion has already started