All Stories — Page 38

Founder John is pitching a multimodal model family around inference speed and structured outputs; Rysana's comparisons use its own benchmark and mid-August API measurements.

A third-party tracker labels Juno an OpenAI-branded ChatGPT plugin, while OpenAI's recent voice update offers a software link, not a hardware confirmation.

Sources familiar with OpenAI's finances told Axios the annualized pace rose more than 70% since July, while the company has not disclosed the costs behind it.

Founder and former surgeon Tom Kelly is extending Heidi beyond visit notes; the rollout starts September 29th in English and French, excluding the UK and EU.

Founder Xiao Hong is extending Manus from an agent that completes work into a creative tool for building, revising and sharing browser games.

Founder Eugenia Kuyda is putting friends and task completion at the center of Wabi, which raised a $20M pre-seed led by Andreessen Horowitz in 2025.

The move would take Cloudflare from relying on outside certificate authorities to seeking a role in the trust infrastructure behind public HTTPS.

The expansion adds connectors for tools including QuickBooks, Shopify and Slack, with Meta saying users must approve anything the agent publishes, sends or spends.

The Medal spinout names six backers and opens an early-access list for models trained on gameplay video paired with controller actions.

Developer Ian Nuttall flagged the label on September 29th, as OpenAI's docs keep separate limits for ChatGPT, Work and Codex.

Md Ismail Sojal promoted the Gemma 4 fine-tune on September 29th; its model card reports no formal benchmarks and warns users to verify its conclusions.

GitHub Security Lab says its open-source workflows found and reported 24 flaws. The examples include an OsmAnd location-tracking bug and a Wikipedia app deep-link issue; running the mobile audit requires a Copilot license and premium model requests.

In announcements published September 28th, Gimlet Labs and Cerebras said their integrated system was already serving inference traffic in private deployments. The first Cerebras-powered Gimlet Cloud data center is expected later in 2026, with customer access to Cerebras' CS-4 system planned for 2027.

The new app lets users group agents to divide tasks; Cue is free in invite-only early access, with iOS availability pending App Store review.

Revenue reached nearly $4.6B in 2025, but a $34B accounting charge drove most of the reported net loss as Anthropic targets a valuation above $2T.

The model was due to debut in ChatGPT and Codex in October, after researchers raised concerns during internal testing, the Journal reports.

Sam Altman's September 29th keynote will most likely feature an always-on assistant, but OpenAI has confirmed no product announcements, it just keeps leaking stuff that implies it. We take a look at all the rumors and leaks for DevDay 20206.

The site promises server-checked moves and spectator explanations, but publishes no match record or test setup that would make a winner meaningful.

The bots can collaborate in Slack and use team-approved tools, while each worker's conversations stay private.

Li will join AMD as executive vice president and chief scientist, while World Labs' research team is set to continue inside the chipmaker.

The feature is available through Fish Audio's transcribe-1-pro API at $0.36 per audio hour; the company says its speech-recognition models support 83 languages.

Co-founder Daniel Tang's AutoTrust AI added a compact decision block to a frozen Qwen3.8-27B model, keeping typed decisions and ordinary generation on one set of weights.

The model is listed at $2 per million input tokens and $10 per million output tokens; Anthropic says it runs 30%+ faster and costs up to 30% less per task.

Founders Jeff Bonwick and Mike Shapiro began iodyne to bring enterprise-style storage to creative teams; the SEC filing names two investors but not their identities.