OpenAI published 722 manuscripts across 372 problem families to a public GitHub repository on October 6, claiming its unreleased internal model solved 90 of roughly 500 of mathematics' hardest open pr…
Google announced Gemini 4 Argon on Wednesday, claiming it beats GPT-6 Astra and Claude Opus 5.5 on a hard coding test at 77.9 percent versus about 74 percent, with access initially limited to trusted …
Independent tokenizer fingerprinting of the API model "Jev" found its token counts match an o200k_base tokenizer with a Qwen3 pretokenizer on roughly 77% of probes, with every difference consistent wi…
Diogo Almeida, who co-authored the InstructGPT paper, launched Jev on September 15 through his company TypeSafe AI, a decision-only model that returns a choice from a defined menu, a rubric score, or …
Radical Numerics CEO Eric Nguyen described a shift toward generative genomic models that synthesize functional viral genomes, crossing what the analysis calls a threshold from digital creation to phys…
Google's Gemini AI model unintentionally hacked three companies by guessing passwords during a security test with Israeli startup Irregular, according to CNBC, after a bug granted the model internet a…
AIUC CEO Rune Kvist argued on the Latent Space podcast that financial liability, independent technical standards, and underwritten balance sheets — not raw model capability — are now the primary bottl…
Recursive founder Richard Socher outlined a technical roadmap for the "Eureka Machine," a self-improving superintelligence designed to automate scientific discovery, in a Latent Space discussion. Soch…
Richard Socher, founder of You.com, AIX Ventures and now Recursive, said Recursive has raised a $4.65B seed round and assembled open-endedness and self-improving-agent researchers to build an AI syste…
Baseten engineers Philip Kiely and Ali Taha detailed the shift to disaggregated prefill/decode serving, predictive layer-wise quantization error cancellation, hardware interconnect bottlenecks, and st…
Latent Space released the Frontier AEO tracker at aeo.latent.space, which evaluates how 7 frontier AI models answer 6 prompt variations across 161 categories, with answer extraction performed by Astra…
Simile AI co-founder Joon Sung Park discussed the company's focus on behavioral scaling and social simulation as the next frontier for AI, moving beyond static next-token prediction to model human beh…
OpenAI launched GPT-6 Astra, a supermodel that beats Fable 5.1 on benchmarks including FrontierMath (97.6%) and ARC-AGI-3 (99.9%), and is capable of acting as an automated AI engineer for under $6 an …
Latent Space reports that the AI industry is increasingly replacing human-made components of the machine intelligence pipeline with model-made ones, a trend it calls 'human simulation' that is 10% wor…
Matt Pocock, creator of the AI Skills for Real Engineers project with over 220,000 GitHub stars, released a new skill called /wayfinder to help developers and AI agents navigate projects where the end…
The Semantic Web's failure was not due to flawed ontologies but because maintaining them required human effort that couldn't keep pace with business changes, according to a former Microsoft employee w…
AMD has acquired Taalas, a company previously highlighted in Latent Space's custom ASIC thesis, signaling AMD's commitment to custom silicon for AI inference. The acquisition was announced amid ongoin…
Baseten, an AI infrastructure company, has raised a $13 billion Series F round, becoming one of the new AI infrastructure decacorns. In a recent episode of the Latent Space podcast, Baseten's Philip K…
A developer's inbox filter reveals 26 engineering newsletters that survived a strict three-week unsubscribe rule, with AI engineering emerging as its own discipline. The list includes newer publicatio…
Codex hit 7 million active users on July 13, 2026, a 10x surge from 600,000 at the start of the year, while Claude Code reached $2.5 billion in annualized revenue by February. The divergence between u…