OpenAI reports that full-stack advances across chips, compute infrastructure, model efficiency, and product packaging have cut AI intelligence costs, with transformer models achieving a 1000x efficien…
OpenAI's head of data centers has left the company, creating leadership churn in the function responsible for turning capital and power access into usable training and inference capacity. The departur…
OpenAI's custom inference chip, Jalapeño, is delivering industry-leading speed and efficiency in AI inference, achieving 2.5x faster inference speeds and reducing power consumption by 40% compared to …
KVBoost, a new inference optimization method detailed in an arXiv paper (arXiv:2608.21362), cuts time-to-first-token on Qwen2.5-3B by 4.49x, from 639.1 ms to 142.4 ms, with no accuracy loss. The techn…
An Amazon Fire HD 10 owner spent $266.15 in LLM usage to root a 2021 tablet in one day, with Kimi K3 identifying that Amazon's firmware still contained the known Mali GPU bug CVE-2022-38181, GLM-5.2 c…
A new arXiv paper (2608.20341) formalizes Spec-Driven Agentic Development (SDAD), a methodology for AI-native software development lifecycles. The paper argues that with coding agents capable of 100K-…
Ox Alpha, a free stealth AI model optimized for coding, sustained agentic work, and production workloads, appeared on OpenRouter with anonymous creators, according to TechCrunch. The unknown provenanc…
PrimeAgentOrchestrator (PAO), a production wrapper for Claude Code agents, pre-loads coding agents with relevant personal memories from PostgreSQL and Cloudflare Worker backends, reducing redundant wo…
Anthropic's Opus 5 model accounted for only $1 million in corporate spend in July 2026, according to billing data from 70,000 Ramp customers, as cheaper models gained favor. The data highlights a tren…
Award-winning Hollywood writers and directors are training AI models in screenwriting and production through lucrative reinforcement learning gigs, according to The Guardian. This injection of elite, …
Hugging Face tests on 11 open-source automatic speech recognition models found several reproduced benchmark transcripts verbatim even when the audio contradicted them, exposing a 10–20% overstatement …
A single LDG.E instruction on an NVIDIA RTX 4090 takes 15 nanoseconds when served from L1 cache, 127 nanoseconds from L2, and 255 nanoseconds from DRAM, revealing a 17× latency gap for memory-bound wo…
OpenAI's new Apple Messages plug-in runs locally on-device to grant ChatGPT direct read, search, write, and delete capabilities over a user's native iMessage and SMS databases, according to TechCrunch…
OpenAI announced Zero Data Retention for eligible API customers, guaranteeing that input and output data from frontier models is not persisted, alongside a new Private Safety Processing feature that r…
Stampli, an accounts payable automation company, reduced launch production hours by 68% by integrating OpenAI's ChatGPT Work and Codex into its delivery pipeline, compressing weeks of release preparat…
ChatGPT search queries using the site: operator jumped from under 0.5% to 16-17% of all search fanouts following the GPT-5.6 rollout, according to Promptwatch tracking data cited by Simon Willison. Th…
A new benchmark, FinSkillBench, shows that equipping financial agents with curated procedural skill packages raises mean task performance from 0.366 to 0.528 on investment tasks, while self-generated …
OpenAI is previewing Private Safety Processing, a feature that monitors for long-horizon API abuse across multiple distinct sessions without retaining any customer conversation data, enabling enterpri…
Replit launched Free Mode powered by OpenAI's GPT-5.6 Luna, eliminating per-token billing for users and allowing them to build small agents and prototypes at zero marginal cost. This move breaks the p…
PINE64, an open-source hardware developer, has halted all production of its Linux-based devices, citing AI-driven component shortages and price spikes of 3–5 times. The company will pause manufacturin…