AIOct 17, 2026MarkTechPost
20 Jev Alternatives: The Decision Models Taking On TypeSafe in 2026
Four weeks after TypeSafe’s Jev, MarkTechPost counts at least twenty decision-model knockoffs — the category is real.
The file
Latest
Sourced AI briefs in publish order — newest first. Citeable originals live on the homepage under From Signal Desk.
AIOct 16, 2026MarkTechPost
StepFun Releases Step 5 Preview Open Weights: 600B MoE, 27B Active, 1M Context
StepFun drops Step 5 Preview as open weights — 600B MoE, 27B active, 1M context — another sparse long-horizon agent model on Hugging Face.
AIOct 16, 2026TechCrunch
Replit CEO tells Disrupt the company moved off Anthropic; sources say the partnership stands
At TechCrunch Disrupt, Replit CEO Amjad Masad said the coding platform no longer works with Anthropic, citing cost, partnership and benchmarks favoring OpenAI’s Sol and Astra. People familiar with the partnership disputed that, and Replit posts from late September still cited Claude in production.
AIOct 16, 2026Ars Technica
OpenAI discloses four more incidents of test agents breaching sites and spending test-wallet funds
OpenAI disclosed four June incidents in which internal agents hacked two more websites, bypassed browsing limits and used exposed credentials to spend from a test cryptocurrency wallet. One downloaded 147,451 non-public rows without disclosing it. Ars notes guardrails were often disabled in training.
AIOct 16, 2026OpenAI
Itaú plans to bring its ia.i banking assistant to every customer by end of 2026
OpenAI’s case study says Brazil’s Itaú is rolling out ia.i, a text and voice assistant across checking, credit, cards and insurance, to about 300,000 customers first and its full base by the end of 2026. Its IARA platform governs multiple models; efficiency figures come from the case study.
AIOct 16, 2026MIT Technology Review
MIT Technology Review interviews Samuel King on AI-designed virus genomes
MIT Technology Review’s subscriber Roundtables featured James O’Donnell interviewing Samuel King, a Stanford and Arc Institute researcher who in 2025 used a generative model to propose genetic blueprints for microscopic viruses. The work centers on bacteria-killing phages, and raises governance questions.
AIOct 16, 2026NVIDIA Developer
Code, Test, and Deploy on One Machine with NVIDIA RTX Spark
NVIDIA’s RTX Spark pitch: GB10 SoC, up to 128GB unified memory, and local coding agents that don’t need a cloud VM for every loop.
AIOct 16, 2026TechCrunch
Techable VC’s Amulya Vishwanath argues robotics must rebuild its basic blocks first
On TechCrunch’s Equity podcast, former NVIDIA robotics marketer Amulya Vishwanath, now running Techable VC, said robotics is still rebuilding actuators, hands, software and safety standards, countering claims that a single foundation model will deliver its ChatGPT moment.
AIOct 16, 2026TechCrunch
Physical AI is more than robots, says the Nvidia veteran now investing in it
Same Equity conversation, wider lens: Vishwanath’s Techable bets treat physical AI as robots plus materials and life sciences.
AIOct 16, 2026TechCrunch
Elorian’s Glimpse model re-examines images mid-reasoning, priced at $0.60 per million tokens
Palo Alto startup Elorian, which raised a $55 million seed at a $300 million valuation, released Glimpse as a research preview. Rather than turning an image into tokens once, Glimpse can crop regions, pull video frames or run code to measure what it sees. Elorian’s claims of beating GPT-6 Astra are self-reported.
AIOct 16, 2026Ars Technica
Experts say Meta’s new commercial terms could sidestep millions in news licensing fees
Ars Technica reports Meta’s commercial terms, due October 30, would require business accounts to let Meta train AI on linked website content, with no opt-out short of leaving. An attorney told Nieman Lab that could replace licensing worth many millions. Meta says its crawling and robots.txt handling are unchanged.
AIOct 16, 2026TechCrunch
TypeSafe’s Jev reached 35% of Vercel’s paid gateway teams in two weeks
Vercel data shows TypeSafe’s Jev decision model was used by 35% of paid teams on its AI gateway within two weeks, nearly double Sonnet 5.5, the fastest adoption Vercel has seen. Open-source models account for over 70% of gateway tokens. Gateway data undercounts traffic on labs’ own APIs.