Explore our reporting on how AI and robotics change work, business and the world.
Build a manual coding-agent skill that collects dated HTML cards with a SQLite cursor, bounded parallel fetches and ordered commits. Copy the complete Python script and skill, then check the JSON result and resume behavior.
Build a local Gemini 3.1 Flash Live voice session that sends combined-desktop frames, checks spoken answers against the screen and stops capture.
Reflection released Beam's BF16, FP8 and NVFP4 weights on October 16. Developers can also use its public-beta API, whose free access has daily limits.
On 51 public EdgeBench tasks, SoL-Pi's authors report 44.7–49.0% less token traffic and about one-third lower estimated API cost than Pi. At xhigh, the full stack's mean score fell from 44.833 to 42.003 with GPT-5.6 Sol and from 44.756 to 42.224 with Opus 5.
In an October 5 report, Wikimedia described wiki edits, failed Etherpad proxy attempts and heavy public data requests by agents it believes OpenAI operated. It found no evidence of compromise.
Plan a multi-shot AI film in Claude, make Nano Banana Pro and Seedance 2.0 shots in Luma, test their actual files in Resolve, and check the finished film's full cost.
Build a complete Claude Code project manual from verified repository facts, review its exact saved version, and check it in a new session.
Use Gemini's Pro and Full music controls to prompt an original structured song, then listen to and reopen the downloaded MP3 before a human rights decision.
Use Claude Cowork to draft a source-checked film office, then deliver approved TXT and CSV copies through private project knowledge and verify every recipient file.
OpenAI’s Decisions API is generally available with JSON inputs, question images and caching. Its typed answers, prices and limits define what builders can evaluate now.
Continues OpenAI DevDay 2026: launches, availability and what builders can use
A wearable system pairs video, trackers, gloves and touch across 100 hours of human activity. Its full-body motion is reconstructed, and robot replay remains a two-clip simulation.
Signed October 22 and effective December 21, the law ties disclosure to false impressions about goods or services and expressly excludes content limited to text.
A community fork of OpenAI’s mathematics collection is recording manuscript edits, reviews and follow-up work. Its growing notes do not certify the proofs.
Continues From unit distances to Navier-Stokes: AI mathematics reaches a new phase
Configure a Paperclip CEO, researcher, marketer and designer around one owned campaign, approve research before creative work, and check the task and run records before using the result.
Inspect and install an exact Journey kit release, check its first use, then package a workflow you have already tested and verify that a teammate can install your release.
WiFireNet detected a concealed, deactivated handgun in controlled room recordings. Its 72.5 handgun AP@25 score and sample-level split leave real screening performance open.
A Nielsen Norman Group survey of 521 UX practitioners finds faster task ratings alongside concerns about AI output and work. The study did not measure completed-work productivity.
The AI-chip developer reported sharply higher sales and profit, alongside negative quarterly operating cash flow and rising inventory.
