The tech world, aggregated.

🤖

AI · 2 hours ago · 6 min read

Frontier models are getting cheaper to run — what that means for startups

Inference costs have dropped 40% year-over-year. We break down the data and what founders should be building now that the economics have shifted.

via TechCrunch

Latest in AI

• AI Open-source multimodal model matches proprietary benchmarks 11h
• AI EU publishes final guidance on model training data disclosure 13h
• AI Why agents need better memory before they can replace workflows 15h
• AI Chip startup claims 3× inference efficiency with new architecture 18h

Startups & Funding

• Startups Fintech infra startup raises $60M Series B led by a16z 1d
• Startups The quiet rise of B2B marketplaces in emerging markets 1d
• Startups YC batch breakdown: what the latest cohort is building 1d

Developer Tools

• Dev Postgres 18 released with async I/O by default 1d
• Dev A visual guide to the new CSS anchor positioning API 2d
• Dev Why your CI is slow: five common pipeline antipatterns 2d
• Dev Bun 2.0 benchmarks: is Node still the default? 2d
• Dev TypeScript 6.0 lands with native project references 3d

Gadgets & Hardware

• Gadgets Framework's modular laptop adds a RISC-V mainboard option 2d
• Gadgets E-ink monitors: now actually usable for coding? 3d
• Gadgets The best budget mechanical keyboards, tested 3d
• Gadgets Smart home Matter update fixes long-standing pairing bugs 4d

Science & Future Tech

• AI Quantum error correction passes a key milestone 3d
• Dev Fusion startup claims net-energy milestone in peer review 4d
• Startups Space cargo startups are betting on monthly launches 4d

Load more stories