AI — Page 11

Founder Jozef Gherman's internal test says 89 of 100 rewrites passed Pangram V4, while 18% failed StealthGPT's factual-consistency check.

The pretraining researcher worked at OpenAI and Anthropic, then concluded that private labs cannot safely coordinate a slowdown.

The open-weight DiffusionGemma fine-tune uses 4B active parameters, while Thesys' own benchmark leaves usability untested.

OpenAI released a 165-page paper and Lean files, while Clay still lists the $1M Millennium Problem as unsolved.

The September 11th event frames cooperation around models, infrastructure and intellectual property, with compensation and rights protection on the agenda.

Mini and Pro target multimodal computer use, while text-only Max requires far more accelerator capacity and Pro remains unavailable for download.

The former DeepMind researcher is moving his Dreamer world models from Minecraft into humanoids, while keeping Embo's product and financing under wraps.

Jinho Jang's project made the modified model downloadable and reported compliance with every HarmBench prompt in its own test. The result has no established independent validation.

The invite-only iPhone app connects to email, calendars, health and financial data, then asks for approval before acting.

Launched in late May with $25M at a reported $225M valuation, NavigateAI uses phones and Meta glasses to coach workers as construction confronts a widening labor shortage.

The San Antonio station says reporters will remain on screen and humans will approve the synthetic voiceovers before broadcast.

Google DeepMind design lead Ammaar Reshi showed Skyrim, Arkham City, Hades and Age of Empires II with touch controls and no streaming.

The demo uses Astra as coordinator while Seedream 5 Pro and Seedance 2.5 generate mockups and video.

Dhravya Shah's memory startup is pushing beyond retrieval as new benchmarks expose how little today's agents retain from experience.

A user test suggests OpenAI's image-only input can infer audio from visual frequency patterns, four days after Astra's release.

Existing users must change the model name, while the $0 endpoint remains subject to unpublished usage caps.

TestingCatalog found references to hosted and self-managed environments, skills and plugins ahead of OpenAI's September 29th DevDay.

Alex Reibman's 24-hour GutCheck test ended with five more users and no new revenue. Bottleneck calls it a $447 loss, although its ledger shows a $99.50 decline.

The 1.653-second benchmark used a four-step adapter, warm hardware and no MP4 encoding; H3's weights retain a separate community license.

The MIT-licensed tool syncs rules, skills, documentation and hooks across competing AI coding tools from a shared repository.

Timothy Gordon's open-source tool preserves project decisions across Claude Code, Cursor, Windsurf and Google Antigravity sessions.

DeepLearning.AI's founder recommends testing, architectural review and calibrated autonomy instead of leaving coding agents unattended.

Mathew Ingram says Ryan Merket's one-person newsroom has revived the old bloggers-versus-journalists fight, with speed no longer reserved for humans.

Jakub Pachocki wants frontier labs to set shared triggers for slowing development, three days after GPT-6 Astra and following an OpenAI agent containment failure.