TrendPulse

https://preview.redd.it/tjbwkmn4djgh1.png?width=1489&format=png&auto=webp&s=d11ec03569d082cdaf806c131b5be19e407187dd How?

Absolute cinema

https://api-docs.deepseek.com/updates/ Its API rn only, but like always I think it will be open weights.

Example chat: https://claude.ai/share/887a2e86-a675-44dd-b7c3-c43211f7ee4c

https://www.wsj.com/tech/banks-in-talks-to-lend-15-billion-for-anthropic-data-center-backed-by-google-606d7afd?mod=author\_content\_page\_1\_pos\_1

"🔷 We’ve massively upgraded its Agent capabilities—benchmark scores are now far surpassing the V4-Pro-Preview. Check out the massive performance leap below! 👇 🔷 The official V4-Flash now natively supports the Responses API format and is fully adapted for Codex Check out the configuration details in our official API docs: https://api-docs.deepseek.com/quick\_start/agent\_integrations/codex/ "

Graph taken from their price cut announcement: Advancing the price-performance frontier with GPT-5.6 | OpenAI

https://openai.com/index/advancing-the-price-performance-frontier-with-gpt-5-6/

It was reportedly running for about 12 hours on Ultracode using a multi-agent loop. Tweet: @Paulius Code: pallet-town-3d

This chart is often used to argue that machines do not replace jobs, because bank teller numbers grew even after ATMs were introduced. But somehow, it is almost always shown without per-capita numbers. So I added them.

Whose buying? Are you buying both? Will one destroy the other? Will they pull a space x and rise before cratering?

Mine: GPT-4: proving pretraining and scaling worked Claude 3.5 Sonnet: proved agentic coding is the future o1: proved test time compute worked o3: one of my biggest oh shit moments along with GPT-4, proved test time compute can scale like crazy Claude Fable: proved that a huge model + test time compute works really well, parameter scaling is nowhere near dead