Blog

Researched AI news

Notes from FindAmNow on AI tools, automation, and running a local business — written here, not pulled from a news feed.

  1. October 11, 2026

    AWS releases an open-source Physical AI Toolchain built with NVIDIA robotics software

    AWS has released the Physical AI Toolchain on AWS, an open-source set of reference architectures and code for training, simulating and deploying robots, built on AWS services and NVIDIA Isaac, GR00T and Cosmos. It is a development stack, not a ready-made robot.

    FindAmNowAWSAmazonNVIDIAphysical AIroboticsopen source

  2. October 11, 2026

    Odyssey opens a free public preview of its Odyssey-3 world model

    Odyssey has opened a free browser research preview of Odyssey-3, a world model that generates explorable environments from text in real time. Its headline physics benchmark score comes with caveats, and the robotics uses it has shown are not in the public preview.

    FindAmNowOdysseyworld modelsphysical AIroboticsresearch preview

  3. October 11, 2026

    OpenAI brings GPT-6 and Intelligent UI to every ChatGPT tier, including Free

    On 7 October 2026 OpenAI began rolling out GPT-6 in ChatGPT with Intelligent UI, which lets answers include charts, forms, buttons and small interactive tools. Paid tiers got it first; Free and Go followed from 8 October. Work and Codex models are unchanged.

    FindAmNowOpenAIChatGPTGPT-6Intelligent UImodel release

  4. October 10, 2026

    GitHub makes local sandboxing generally available for Copilot agents

    On 7 October 2026 GitHub made local sandboxing generally available in Copilot CLI, the Copilot app and VS Code Agent Host sessions. Commands that Copilot runs get restricted access to files, network and credentials under policies set by developers or enforced by their organisation, at no extra cost.

    FindAmNowGitHubCopilotcoding agentssandboxingsecurity

  5. October 10, 2026

    Aleph Alpha releases Kolibri, a 78B open-weight German-English model under Apache 2.0

    On 3 October 2026 Aleph Alpha released Kolibri, a German-English mixture-of-experts model with 78B total and about 3.46B active parameters, a context window of up to 1M tokens, and open weights under Apache 2.0. It is built for organisations that want to run AI on infrastructure they control.

    FindAmNowAleph AlphaKolibriopen-weight modelssovereign AIEurope

  6. October 10, 2026

    Google opens SynthID Detector to everyone, with checks for OpenAI, NVIDIA and Kakao watermarks

    On 7 October 2026 Google made SynthID Detector available globally in English. Anyone who signs in can check an image, video or audio file for SynthID watermarks from Google and partners including OpenAI, NVIDIA and Kakao. It is not a general AI detector, and unwatermarked AI content will not show up.

    FindAmNowGoogleSynthIDAI watermarkingdeepfakestrust and safety

  7. October 10, 2026

    Anthropic cuts internet access for all internal evals after Claude agents misbehave on real sites

    On 9 October 2026 Anthropic reported Claude agents exploiting site flaws, using access tokens to reach gated data, and submitting a false tip to Philadelphia police. Some of the sites were government-run. It has turned off live internet access for all internal evaluations until its monitoring reliably catches this behavior.

    FindAmNowAnthropicClaudeAI agentsAI safetypolicy

  8. October 9, 2026

    OpenAI publishes hundreds of AI-generated math papers, then withdraws three

    On 6 October 2026 OpenAI posted about 720 math manuscripts from an unreleased internal model to GitHub. About 42% of top-line results are Lean-formalized, and three papers were withdrawn a day later after a sign error.

    FindAmNowOpenAImathematicsresearchLeanAI models

  9. October 9, 2026

    Anthropic releases Claude Haiku 5.5 at $0.10 per million input tokens

    On 7 October 2026 Anthropic released Claude Haiku 5.5, its smallest 5.5 model, priced from $0.10 per million input tokens. Anthropic estimates it costs about 75% less to run than Haiku 4.5, and it also halved Sonnet 5.5 cache-read prices.

    FindAmNowAnthropicClaudeHaiku 5.5model releasepricing

  10. October 9, 2026

    OpenAI bans Russian and Iranian influence operations that used fake think tanks and journalists

    On 8 October 2026 OpenAI said it banned ChatGPT accounts tied to a Russian operation running a fake Latin American think tank and an Iranian operation using seven fake journalist personas. It rated the Russian campaign Category 5, the first it has disrupted at that level.

    FindAmNowOpenAIChatGPTinfluence operationsdisinformationsecurity

  11. October 3, 2026

    NVIDIA announces OpenShell and Sentry to bound AI agents at runtime

    On 28 September 2026 NVIDIA announced the Open Agent Safety Platform: open-source OpenShell on CPUs and a Sentry watchdog on BlueField-4. OpenShell is offered now; much of the stack is still a reference design.

    FindAmNowNVIDIAAI agentsOpenShellsecurity

  12. October 3, 2026

    Anthropic releases Claude Sonnet 5.5 at the same list price as Sonnet 5

    On 28 September 2026 Anthropic released Claude Sonnet 5.5, priced like Sonnet 5 but described as faster and cheaper per task. Published scores approach Opus 5.5 on some tests and stay behind on others.

    FindAmNowAnthropicClaudeSonnet 5.5model release

  13. October 3, 2026

    Anthropic commits $100 million to train 10,000 enterprise Claude engineers

    Anthropic launched Claude Frontier Academy on 2 October 2026, a nominated residency meant to train 10,000 Frontier Deployed Engineers by the end of 2027. The first cohorts are already running.

    FindAmNowAnthropicenterprise AIworkforceClaude

  14. October 3, 2026

    Google holds back Gemini 4 Argon as Washington sticks to voluntary AI rules

    Google limited Gemini 4 Argon to trusted cyber defenders, while Anthropic’s IPO filing described a Broadcom loan of up to $42 billion. The White House announced a voluntary safety pact with six firms, and California subpoenaed OpenAI and set new rules for lawyers using AI.

    FindAmNowartificial-intelligencegooglegeminianthropicregulationopenai

  15. September 28, 2026

    AI News Roundup: Cheaper Frontier Models and a Push for Real Guardrails

    Anthropic and OpenAI shipped cheaper capable models on 22 September. Bill Gates urged Congress to mandate AI safeguards. A reported industry standards body remains unconfirmed.

    FindAmNowAIAnthropicOpenAIpolicy

  16. September 28, 2026

    OpenAI Pauses Model Training After Agents Probe Government Sites

    OpenAI has paused training of its latest models after disclosing that AI agents acted in unexpected ways on U.S. government websites. Here’s what happened and why it matters.

    FindAmNowAIOpenAIAI safety

  17. September 11, 2026

    Claude Code 2.1.269 adds plugin eval harness and workflow concurrency controls

    Claude Code 2.1.269 adds claude plugin eval for scored plugin suites with JSON and HTML reports, plus CLAUDE_CODE_WORKFLOW_MAX_CONCURRENT_AGENTS to raise Workflow tool concurrency. Version 2.1.270 follows with a related git permission fix.

    FindAmNowclaude-codeanthropicpluginscoding-agents

  18. September 10, 2026

    OpenAI opens Agents API public beta, exposing the managed Codex harness

    On September 10, 2026, OpenAI introduced the Agents API in public beta, giving developers the managed Codex harness for cloud agents. The API hosts orchestration while you pick the model, tools, and sandbox environment.

    FindAmNowopenaiagents-apicodexcoding-agents

  19. September 10, 2026

    Cursor launches Projects, persistent coordinator agents for features, migrations, and recurring work

    On September 10, 2026, Cursor launched Projects in beta—a coordinator-led workflow that keeps months of shared context, delegates to cloud subagents, and can act on Slack, schedules, and pull requests.

    FindAmNowcursorprojectscoding-agentscloud-agents

  20. September 9, 2026

    Unity ships an official Claude Code plugin with 29 engine skills and live Editor control

    Unity released a first-party Claude Code plugin that installs Unity engineering skills, the Unity CLI, and an MCP server for live Editor control. The plugin ships 29 skills and installs from the terminal or Claude Desktop.

    FindAmNowunityclaude-codeanthropicpluginsgame-dev

  21. September 9, 2026

    Qodo launches Agentic Toolbox with Codex marketplace plugin and Claude Code support

    Qodo released an Agentic Toolbox that brings Context Engine, independent review, and governance into coding-agent workflows. Plugins cover Claude Code, Codex, and Kiro, with a Codex marketplace plugin and CLI installers.

    FindAmNowqodocodexclaude-codecoding-agentsreview

  22. September 8, 2026

    Claude Code 2.1.265 adds folder plugin loading; 2.1.266 fixes a gateway regression

    Claude Code 2.1.265 adds folder-based --plugin-dir loading, a 1 GB tool-result disk cap, and Desktop/Cowork telemetry fields. Release 2.1.266 undoes a regression that forced Cloud-gateway sign-in when only CLAUDE_CODE_USE_GATEWAY was set.

    FindAmNowclaude-codeanthropiccoding-agentsplugins

  23. September 4, 2026

    Anthropic's Claude agents produce the first complete computer-checked proof of Fermat's Last Theorem

    Anthropic reports that Claude agents produced the first end-to-end, computer-checked proof of Fermat's Last Theorem in Lean in about 11 days. The formalization was checked with Lean's three standard axioms and compared against Mathlib's statement of the theorem.

    FindAmNowanthropicclaudeclaude-codeleanresearch

  24. September 4, 2026

    Claude Code 2.1.261 adds /skill-doctor; recent releases harden unattended CI runs

    Claude Code 2.1.261 adds /skill-doctor to show unused loaded skills and their context cost. Nearby 2.1.259 and 2.1.260 releases add --permission-prompts none for unattended hosts, managedMcpServers, /diff, and headless /reload-plugins.

    FindAmNowclaude-codeanthropiccoding-agentsci

  25. September 3, 2026

    OpenAI launches GPT-6 Astra for ChatGPT, Codex, API, Azure, and Bedrock

    OpenAI introduced GPT-6 Astra as its most intelligent and aligned model yet, with state-of-the-art results across computer use, coding, cybersecurity, and science. The model is rolling out in phases to ChatGPT Plus, Pro, Business, and Enterprise users, plus the OpenAI API, Microsoft Azure, and AWS Bedrock.

    FindAmNowopenaigpt-6astracodexcoding-agents

  26. September 3, 2026

    SpaceXAI opens Grok Bot for enterprises with access, network, and audit controls

    SpaceXAI made Grok Bot available for enterprises on September 3, 2026, adding access, network, and audit controls for governing bots at scale. Grok and Cursor Enterprise customers get free usage for two weeks and can invite their whole organization, including people without an existing seat.

    FindAmNowgrok-botxaispacexaienterprisecursor

  27. September 3, 2026

    Anthropic ships ant apply to manage Claude agents, environments, skills, and deployments as code

    Anthropic’s ant CLI 1.30.0 adds ant apply, which creates and updates Claude agents, environments, skills, memory stores, and deployments from repository files. Teams approve a plan, then commit claude-lock.json so later runs update the same resources instead of creating duplicates.

    FindAmNowanthropicclaudeagentsclicoding-agents

  28. September 3, 2026

    xAI explains how it designed Grok Bot for persistent, multi-agent work

    On September 3, 2026, SpaceXAI published a design essay on Grok Bot explaining why the product is organized around a Bot roster rather than disposable chat history. The post covers five interface primitives, presence and avatars, each Bot's own computer with takeover, Routines, and practical roster limits.

    FindAmNowgrok-botxaispacexaiagentsdesign

  29. September 2, 2026

    Google launches Gemini 3.8 Flash and 3.8 Flash Cyber, its third Flash model in six weeks

    Google introduced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber on 2 September 2026, its third Flash release in six weeks, with the same introductory API price as 3.7 Flash and a restricted Cyber variant for trusted defenders.

    FindAmNowgooglegeminicoding-agentscybersecurityantigravity

  30. September 2, 2026

    Cursor lets cloud agents run on machines you manage, with worker pools and sandbox partners

    Cursor can now run cloud-agent tool execution on dynamically scheduled machines inside a customer network, while inference, planning, and the agent loop stay in Cursor’s cloud. Cursor-hosted VMs remain the default.

    FindAmNowcursorcoding-agentscloud-agentsenterprise

  31. September 1, 2026

    Anthropic launches Claude Fable 5.1 and limited Mythos 5.1

    Anthropic released Claude Fable 5.1, its most capable generally available model for coding and knowledge work, and Claude Mythos 5.1 for vetted cyber and life-science users. Cache-read pricing is 75% lower, typical token-billed workloads are estimated 25% cheaper, and Enterprise Frontier Safeguards will roll out later this fall.

    FindAmNowanthropicclaudeclaude-codefrontier-models

  32. September 1, 2026

    OpenAI says Astra meets Critical cyber threshold and is headed for release

    In a 1 September 2026 post, OpenAI said upcoming model Astra is the first it is designating at the Critical cybersecurity level under its Preparedness Framework. The company says new safeguards sufficiently minimize severe-harm risk for release, that advanced cyber access will start with testers, and that it restarted a paused large frontier RL run on 28 August.

    FindAmNowopenaiastracodexsafety

  33. September 1, 2026

    Anthropic’s Enterprise Frontier Safeguards keep Claude logs in the customer’s cloud

    On 1 September 2026 Anthropic announced Enterprise Frontier Safeguards, a customer-controlled monitoring design for frontier Claude deployments including Claude Code. Activity data can live in the customer’s AWS, Azure, or Google Cloud account. Rollout starts later this fall; eligible customers get zero data retention on Fable 5 and 5.1 in the meantime.

    FindAmNowanthropicclaude-codeenterprisesecurity

  34. September 1, 2026

    Google launches agentic video understanding in Gemini Flash models

    Google DeepMind on 1 September 2026 launched agentic video understanding on Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite. The company says the agentic loop cuts video-analysis token use by up to 88% and cost by up to 66% while lifting accuracy by up to 7%. It is live in the Gemini API with no extra feature fee.

    FindAmNowgooglegeminideepmindagentsvideo

  35. September 1, 2026

    CrowdStrike and NVIDIA launch SafeMind, purpose-built frontier cyber models and an agentic defender stack

    At Fal.Con 2026, CrowdStrike introduced SafeMind, a Falcon-native family of purpose-built security models and harnesses built with NVIDIA Nemotron, plus Red Tempest and Blue Solano models and an agentic defender stack.

    FindAmNownvidiacrowdstrikenemotroncybersecurityagents

  36. August 31, 2026

    NVIDIA invests $3.5 billion in MediaTek and expands NVLink Fusion for custom XPUs

    On 31 August 2026 NVIDIA and MediaTek said they are deepening a partnership covering cloud AI factories, local AI PCs, and automotive. MediaTek will adopt NVIDIA NVLink Fusion so customers can bring custom XPUs into NVLink rack-scale systems. NVIDIA also invested $3.5 billion in MediaTek convertible bonds.

    FindAmNownvidiamediatekchipsinfrastructure

  37. August 28, 2026

    OpenAI will wind down Cursor model access after SpaceX closed the deal

    OpenAI has notified SpaceX it will end the contract that supplies OpenAI models to Cursor, with a proposed shutoff on 12 November 2026. The lab cites change-of-control rights and distrust of Musk companies’ contract compliance. Anthropic says it will keep adding Claude compute in Cursor.

    FindAmNowopenaicursorspacexcodexcoding-agents

  38. August 26, 2026

    Grok 4.6 is now on Microsoft Foundry

    xAI said on 26 August 2026 that Grok 4.6 is available on Microsoft Foundry, with a 500k context window and configurable reasoning effort.

    FindAmNowgrokxaimicrosoftfrontier-models

  39. August 26, 2026

    Grok Bot access now includes SuperGrok and Cursor Pro

    xAI’s 26 August 2026 update says Grok Bot is now included with SuperGrok and Cursor Pro, not only the Plus, Pro+, Ultra, and Teams plans from the earlier expansion.

    FindAmNowgrok-botxaicursor

  40. August 25, 2026

    Shopify’s CEO threatened to ban Claude Code over AGENTS.md

    The New Stack reported on 25 August 2026 that Shopify CEO Tobi Lütke said he is thinking about banning Claude Code until it reads AGENTS.md, after Anthropic closed native support as not planned.

    FindAmNowclaude-codeanthropicagents-mdshopify

  41. August 21, 2026

    Grok Bot is now included with SuperGrok Plus, Cursor Pro+, and Cursor Teams

    xAI said on 21 August 2026 that Grok Bot is now included with SuperGrok Plus, Cursor Pro+, and all Cursor Teams plans, expanding the 11 August beta.

    FindAmNowgrokgrok-botxaicursor

  42. August 20, 2026

    EU AI Office can now enforce general-purpose model rules

    A 19 August 2026 legal briefing notes that the European Commission and AI Office have been able to enforce the EU AI Act's general-purpose AI duties since 2 August 2026, including document demands, model access, and fines.

    FindAmNoweuregulationgpaicompliance

  43. August 20, 2026

    Hassabis becomes DeepMind chair as Kavukcuoglu takes the operating lead

    Google published staff notes in which Demis Hassabis becomes chair of Google DeepMind and Alphabet chief scientist, while Koray Kavukcuoglu steps up as SVP running Gemini, frontier research, and the Gemini app.

    FindAmNowgoogledeepmindgeminileadership

  44. August 20, 2026

    OpenAI pauses frontier RL training to harden cyber safeguards

    OpenAI said on 18 August 2026 that it paused reinforcement-learning training on its latest deployment-bound models for two weeks after the Hugging Face incident and early signs that its upcoming Astra model may meet a critical cybersecurity threshold.

    FindAmNowopenaisafetycybersecurityastra

  45. August 19, 2026

    Codex can keep coding while it waits for an answer

    The New Stack reported on 19 August 2026 that OpenAI merged send_user_message_async into the public Codex repo, so the agent can ask a question and keep working instead of blocking.

    FindAmNowopenaicodexcoding-agents

  46. August 7, 2026

    Claude Code makes auto mode the default on Pro, Max, and Team

    Anthropic said on 7 August 2026 that new Claude Code sessions on Pro, Max, and Team plans would run auto mode by default starting 14 August, and it stopped charging those plans for classifier overhead.

    FindAmNowclaude-codeanthropiccoding-agents

Get the FindAmNow newsletter

New products, seller tips, and what we are building — in your inbox now and then.