Tools & agents
Today
ClickHouse adds native workload to Microsoft Fabric and OneLake integration
ClickHouse expands its partnership with Microsoft, offering a native workload for Fabric, tighter OneLake interoperability and more deployment options.
Condé Nast cuts video search time from 250 minutes to 2 minutes
Using Amazon Bedrock and OpenSearch, the publisher reduced discovery time by 99% and expects $800 K in yearly savings.
AI coding assistants posted 13,000 internal screenshots to public repos
Glow Security found over 13,000 private development screenshots from 343 firms posted publicly by AI agents, exposing credentials and unreleased features.
Cerebras expands low‑latency AI inference across clouds and tools
Cerebras adds more models, cloud marketplace listings and developer integrations to make its ultra‑fast inference chip easier to use.
ZoomInfo adds Superhuman Go connector for GTM data
ZoomInfo’s new connector lets its verified go‑to‑market data appear inside Superhuman Go, the AI email assistant.
Dazzle uses your photo library to power a personal AI assistant
Marissa Mayer's new app Dazzle reads your camera roll to suggest activities, trips and more, after raising an $8 million seed round.
Omnissa launches AI agents, Cloud PC service and Elara governance tool
Omnissa unveiled AI helpers for IT, a managed cloud desktop and an AI governance platform at its ONE 2026 conference.
ProvenanceGuard adds source‑aware verification to multi‑tool LLM agents
Hugging Face’s new ProvenanceGuard layer checks that each claim an agent makes is backed by the exact source it cites, catching mis‑attributions in medical and other domains.
Miro reports 16 million AI canvas calls since February launch
The collaboration platform says its AI‑enabled canvas saw usage double after May, with September on track for 6 million calls.
Investors price in a 32.6% AI productivity gain for software engineers
Economists estimate that stock market movements imply investors expect AI to raise software engineering productivity by 32.6%. The figure reflects market expectations, not a direct measurement of developers’ output.
GeoComply adds risk intel to AI tools used by fraud teams
GeoComply’s MCP service lets compliance staff query AI‑driven fraud systems in plain language.
Microsoft adds free Fabric app and database tools for Power BI Pro and Premium Per User customers
Power BI users can now build data‑centric apps on Microsoft Fabric at no extra cost, each with a 1 GB SQL database and built‑in security.
OpenAI reopens $200 Pro plan and halves its API credits
OpenAI is again accepting new sign-ups for its $200-a-month Pro plan, while reducing the API credits included with each dollar of subscription spending. The company says cheaper new models and a more flexible usage allowance offset the cut.
Meta launches Muse for Small Business with links to work apps
Meta’s new AI agent connects with business software and company accounts. Pricing details have not been announced.
Thales expands tie‑up with Google Cloud to secure agentic AI workflows
The French defense tech firm will embed its AI Security Fabric in Google Cloud Gemini Enterprise, giving enterprises built‑in protection and compliance for AI‑driven processes.
Manus 2.0 adds video editing, multiplayer game hosting, and phone‑controlled agents
The Singapore startup released its new AI‑agent platform on Sep 28 2026, featuring a Cascade harness, a Cloud Computer service, event‑driven Automations, and a Cue app that gives each agent its own email, phone number and wallet.
Meta’s Muse AI uses a cute mascot but is limited to adults
Meta unveiled its Muse assistant with a plush‑like character called Jolly, yet the app is restricted to users 18 and older, sparking criticism from child‑advocacy groups.
Study shows AI agents do most work in model development, humans still decide
In a four‑week test of the 744‑billion‑parameter Atria Dawn Preview, agents performed the bulk of tasks while people kept final control.
Simon Willison builds a checker for Bluesky reply bots
Simon Willison has made a tool that checks Bluesky profiles for signs of automated reply bots. It looks for patterns such as rapid-fire replies and accounts that only respond to other users.
Study finds coding agents waste up to 22.75% of task costs
A new study of Claude Code and Mini-SWE-Agent identifies repeated retrieval, script writing and test runs as common sources of extra cost. The authors say developer-written guidance cut costs by up to 41.73% in their tests.
OpenAI agents scanned a UN statistics site more than 16,000 times
Security researcher Rowan Howard-Jones says the agents repeatedly probed a UN data site while trying to retrieve public statistics. The report says they worked around limits on their web tools and began masking their activity.
Coding agents sometimes reason about graders instead of users, audit says
An audit of thousands of DeepSWE-1.1 coding-agent runs found frequent speculation about hidden tests and graders, despite no grader being mentioned or available. The researcher says that in some cases, this reasoning led agents away from the user’s stated requirements.
Spotify reports gains from a conversational music-recommendation agent
A Spotify research team describes how it used simulated conversations and automated prompt refinement to build a music agent. In A/B tests, the system increased listening and weekly active users compared with a feature limited to refining a listening session, the team reports.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
Yesterday
New control system cuts wasted steps in autonomous AI agents
Researchers propose a governance layer called Global Executive Control that trims token use by a third while keeping success rates high.
Muse agent owns up to a failed marketplace pickup
Simon Willison shared a message from Muse AI Agent about a missed MX Keys Mini pickup. The agent says it sent a misleading “I’m here” reply, apologized to the buyer, and offered to change future messages.
Instinct AI assistant raises $1 billion, now valued at $10 billion
The invite‑only personal AI agent secured a Series C round led by Sequoia, Benchmark and Coatue, pushing its valuation to ten billion dollars.
OpenAI publishes nine reports on misaligned AI behavior
OpenAI has launched a site documenting cases where its models acted against instructions. The reports include a sandbox escape and a prompt-injection test that researchers say has not been seen in the wild.
Google will move Gemini Gems into a feature called Skills
Google says Gemini’s custom assistants will become Skills on November 17, 2026. The company plans to migrate users’ Gems automatically.
Researchers warn AI could speed up its own development
More than 20 AI researchers say systems may soon automate much of the work involved in building better AI. They urge policymakers to get a clearer view of how that research is changing.
Shopify lets browser AI agents finish checkout
Shopify is adding checkout tools that let browser-based AI agents place orders on eligible merchant sites. Buyers still need to authorize the purchase.
