Latest AI Productivity News

Curated daily - the AI tools, productivity launches, automation updates, and platform releases worth your time. We link straight to the original source.

TechCrunch
AI Tools

TechCrunch

Inherent Releases Faraday, a 27B AI Scientist That Outperforms Claude and GPT at Replicating Research

London-based Inherent, founded by Google DeepMind alumni, released Faraday — a 27B-parameter AI Scientist agent trained with long-horizon reinforcement learning that outperforms Claude Opus 4.8 and GPT-5.5 on Replica, a new 310-task benchmark across 100 ML and AI-for-science papers. The startup emerged from stealth alongside a $50 million seed round led by Index Ventures, positioning Faraday as a stepping stone toward autonomous scientific discovery.

Aug 22, 2026

TechCrunch
Business AI

TechCrunch

Harvard Business School Deploys HeyGen AI Avatars to Coach Founders in Its $699 HBS Foundry Bootcamp

Harvard Business School partnered with HeyGen to put AI avatars of its instructors inside the eight-week, $699 HBS Foundry online bootcamp, where the avatars give entrepreneurs real-time feedback during practice pitches and simulated board meetings without waiting for human instructor availability. The integration marks one of the first uses of AI avatar instructors inside an accredited HBS program, signaling a shift in how top business schools are scaling personalized coaching.

Aug 22, 2026

TechCrunch
AI Tools

TechCrunch

Nvidia Research Shows AI Harness — Not the Model — Is the Real Performance Driver, Lifting Claude Opus 5 to 100% on ARC-AGI-3

Nvidia researchers demonstrated that a custom harness with strong memory management and a supervisor component pushed Claude Opus 5 to a perfect 100% on ARC-AGI-3 — a demanding 2D-game benchmark with no given instructions — a score unachievable without the scaffolding. The finding makes the case that for long-horizon agentic work, the wrapper around a model often matters more than upgrading the model itself.

Aug 21, 2026

Tech Startups
AI Tools

Tech Startups

Twin1 AI Raises $20M Seed to Give Every Knowledge Worker a Personalized AI Digital Twin

Twin1 AI emerged from stealth with a $20 million seed co-led by Bessemer Venture Partners, Tribeca Venture Partners, and Aramco Ventures to build AI-powered digital twins that absorb a professional's emails, meetings, documents, and institutional knowledge, automating 30–50% of their communications work. The twins operate across Slack, Teams, Outlook, Gmail, Google Drive, and SharePoint, making individual expertise scalable across enterprise teams.

Aug 20, 2026

TechCrunch
Business AI

TechCrunch

Ramp Launches Router, an AI Model Routing Service That Cut Its Own Inference Costs 40% Before Going Public

Ramp released Router — an AI model routing service the company quietly ran internally for three years — that automatically sends each API request to the lowest-cost model from OpenAI, Anthropic, DeepSeek, and others that meets the task's performance bar. Early customers have cut inference spend by 40% on average; the service is free through 2026 with a $26 launch credit and is currently US-only.

Aug 20, 2026

Crypto Briefing
Anthropic Claude

Crypto Briefing

Anthropic Opens Claude Academy, a Free Learning Hub Built From Its Internal 4D AI Fluency Training

Anthropic launched Claude Academy — a free public platform teaching safe and effective AI use through the same 4D AI Fluency Framework the company uses for employee onboarding — with courses, tutorials, badges, and personalized recommendations focused on mindsets and problem-first practice rather than feature checklists. The launch extends Anthropic's education push alongside the recently introduced Claude for Teachers.

Aug 20, 2026

Anthropic
Anthropic Claude

Anthropic

Anthropic Reveals Claude Designed Working Protein Binders for 14 of 15 Targets, Beating Industry Hit Rates

Anthropic published lab-validated results showing Claude autonomously designed protein binders against 14 of 15 targets at hit rates of 22–35%, compared to the 10–15% typical in current protein design campaigns, using publicly available specialist models after a single 30,000-token expert prompt. Some of Claude's strongest binders outperformed the best previously published designs, marking a milestone for AI-driven drug discovery.

Aug 20, 2026

VentureBeat
Productivity

VentureBeat

Slack Launches Slack Code, Putting AI Coding Agents in Dedicated Team Channels So Everyone Can Watch, Steer, and Ship

Slack launched Slack Code on August 20 — dedicated project channels where teams can tag in AI coding agents including Claude Code, Cognition's Devin, GitHub Copilot, and Vercel's agent, then review code diffs, get live HTML previews, and approve changes before anything ships, all from within Slack. The feature is available on any Slack plan, though users must bring their own agent credentials.

Aug 20, 2026

9to5Mac
OpenAI / ChatGPT

9to5Mac

ChatGPT Adds Apple Messages Plugin for Mac, Letting Users Read, Search, and Draft Texts From Inside the Chat

OpenAI released an Apple Messages plugin for the ChatGPT macOS desktop app that lets users on all plans read and search their Messages history, then draft or send replies without leaving ChatGPT — compatible with ChatGPT Work and Codex. The same update added read-only conversation sharing from ChatGPT Work and Codex, and the ability to invite workspace members to co-edit Sites.

Aug 20, 2026

PYMNTS
Business AI

PYMNTS

Nvidia Discussions to Invest in AI Data Startup Mercor at $20 Billion Valuation, Double Its Price Nine Months Ago

Nvidia is in talks to invest in Mercor, an AI data annotation company it already pays tens of millions per quarter to source human-expert training data for its Nemotron open-source models, at a $20 billion valuation — double the $10 billion Mercor reached in its Series C just nine months ago. The deal would lock in a strategic supply chain for high-quality data as Mercor crosses $2 billion in annualized gross revenue.

Aug 20, 2026

Help Net Security
OpenAI / ChatGPT

Help Net Security

OpenAI Halts Its Largest Frontier Training Run After Astra Model Flags a "Critical" Cybersecurity Capability Threshold

OpenAI paused its biggest planned frontier reinforcement learning run after internal evaluation of its next-generation Astra model found it may have crossed the "Critical" cybersecurity threshold in its Preparedness Framework — a level that implies potential to autonomously identify and exploit zero-day vulnerabilities in hardened critical systems. The company has spent the interim hardening research environments, expanding monitoring, and running sandboxed evaluations before restarting at reduced scale.

Aug 19, 2026

Search Engine Land
Marketing AI

Search Engine Land

ChatGPT Ads Expand to 31 European Countries Starting August 24 in OpenAI's Largest Global Ad Rollout

OpenAI announced that ChatGPT Ads will begin serving across 31 European markets on August 24, extending to Germany, France, Spain, Italy, the Netherlands, and 26 other countries — its largest global expansion since launching ads in the US six months ago. Ads will appear only to Free and Go-tier users, Pro and Enterprise subscribers stay ad-free, and EU compliance required satisfying GDPR consent rules that delayed the regional launch.

Aug 19, 2026

Anthropic
Anthropic Claude

Anthropic

Anthropic Takes Files API and Admin API to General Availability and Makes Claude Sonnet 5's Introductory Pricing Permanent

Anthropic moved the Files API and Admin API user-management endpoints out of beta on August 19, removing the previously required beta headers for all Claude Enterprise and API customers. In the same update, Anthropic canceled the September 1 price increase for Claude Sonnet 5, locking in the introductory $2 per million input / $10 per million output token rate as the permanent standard price.

Aug 19, 2026

TechCrunch
OpenAI / ChatGPT

TechCrunch

OpenAI Launches ChatGPT for Teens With Built-In Safety Protections, Study Mode, and Parental Controls

OpenAI launched a dedicated ChatGPT experience for users aged 13–17 on August 18, automatically routing minors into teen mode using age estimation, with restrictions on self-harm, romantic, and explicit content, a Study Mode that guides rather than gives answers, and optional parental controls. The release comes as OpenAI faces lawsuits claiming the chatbot contributed to teen mental health crises.

Aug 18, 2026

Anthropic
Anthropic Claude

Anthropic

Anthropic Extends Compliance API to Claude Cowork and Claude Code, Giving Enterprise Security Teams Unified Audit Access

Anthropic expanded its Compliance API in beta for Claude Enterprise customers to cover session content and metadata from both Claude Cowork (desktop, web, and mobile) and Claude Code (CLI and desktop app), letting security and legal teams pull unified logs for audits and eDiscovery through a single interface. The rollout does not yet cover Claude Code on the web or API sessions running on Bedrock, Vertex AI, or Microsoft Foundry.

Aug 18, 2026

TechCrunch
Microsoft Copilot

TechCrunch

Microsoft Merges Its Two Copilot Apps Into One and Retires Group Chats, AI Podcasts, Deep Research, and Its Mico Mascot

Microsoft unified its consumer and business Copilot apps and cut Group Chats, AI-generated podcasts, Copilot Labs, and Deep Research for free-tier users starting August 18, plus retired the Mico animated mascot — moves that follow an internal memo from Copilot chief Jacob Andreou saying the product needed to "earn the right to exist." Paid subscribers get Researcher as a Deep Research replacement.

Aug 18, 2026

LangChain
AI Tools

LangChain

LangChain Releases Deep Agents v0.7 With 65% Token Reduction, In-Context Code Execution, and Per-Model Harness Profiles

LangChain shipped Deep Agents v0.7, cutting baseline input token usage by 65% through harness optimization, and added CodeInterpreterMiddleware for in-context code execution via a scoped QuickJS runtime, DeltaChannel for incremental checkpointing that keeps checkpoint sizes small as threads grow, and HarnessProfile — a per-provider config bundle that auto-applies system prompt tweaks and tool overrides when a model is selected. The update makes running large multi-step agent workflows meaningfully cheaper and easier to tune across providers.

Aug 18, 2026

TechCrunch
Anthropic Claude

TechCrunch

Anthropic's Annualized Revenue Surges to $65 Billion, Nearly Tripling in Three Months Ahead of Expected IPO

Anthropic's annualized revenue run rate hit $65 billion at the end of July — up from $47 billion in May and $9 billion a year earlier — with investors projecting full-year 2026 revenue between $100 billion and $120 billion as Claude adoption accelerates across enterprise. The milestone lands ahead of a widely anticipated IPO and follows Anthropic's first profitable quarter in Q2 2026.

Aug 17, 2026

TechCrunch
Productivity

TechCrunch

Wispr Raises $280M at $2B Valuation to Turn AI Voice Dictation Into a Universal Computing Layer Across Every App

AI voice startup Wispr closed a $280 million Series B led by Menlo Ventures, valuing the company at $2 billion on the back of its Flow product — a speech-to-text layer that works in any text field on any app and has already produced more than 60 billion words across Fortune 500 employees and 10,000 businesses. The new capital will fund an expanded Canto speech model that cuts word error rates from 30% to under 10%, plus a push into meeting notes and broader voice-based computer interaction.

Aug 17, 2026

TechCrunch
Business AI

TechCrunch

Groq Raises $350M and Pivots From Custom AI Chips to Neocloud Targeting 200 MW of GPU Capacity by 2027

Groq raised $350 million at a $3.5 billion valuation — down from $6.9 billion before Nvidia hired its founder and licensed its LPU chip technology — and is reorienting from building proprietary AI chips to operating a GPU-based neocloud with plans to grow from 54 MW to over 200 MW of compute capacity by 2027. The shift positions Groq to compete directly with AWS, Azure, and Google Cloud for AI inference workloads rather than the chip market.

Aug 17, 2026

Axios
OpenAI / ChatGPT

Axios

OpenAI and Nvidia Announce 8-Gigawatt Ohio Data Center With $105 Billion in Nvidia Financing — the Largest AI Compute Site Ever Planned

OpenAI and Nvidia unveiled plans for the Ports-Pike AI data center campus near Cincinnati, Ohio — a former uranium enrichment site that could eventually reach 8 gigawatts of capacity, making it the largest planned AI compute facility in history. Nvidia will invest $1.5 billion in SB Energy, the SoftBank-linked developer, guarantee up to $105 billion in financing, and serve as the sole compute supplier as the first 4.25 GW phase comes online in 2028.

Aug 17, 2026

Google AI for Developers
Google Gemini

Google AI for Developers

Google Retires Imagen 4 API Endpoints Today, Forcing Developers to Migrate to Gemini Image API at Higher Cost

Google shut down the Imagen 4 standard, fast, and ultra endpoints in the Gemini API on August 17, ending access for all three generate model IDs and breaking any app that calls them without a migration. The recommended replacement — Gemini 3.1 Flash Image — is not a drop-in swap and costs up to $0.067 per 1,000 images versus $0.02–$0.06 for Imagen 4, meaning most teams face both a code change and a price increase.

Aug 17, 2026

TechCrunch
Business AI

TechCrunch

Stripe Finalizes $7B+ Acquisition of AI Gateway Startup OpenRouter, Integrating Model Routing Into Its Payments Stack

Stripe has finalized its acquisition of OpenRouter — the AI model router used by 8 million developers to access more than 400 AI models — for over $7 billion, a 5.4x premium over its May 2026 Series B valuation. The deal folds model selection and AI spending directly into Stripe's payments infrastructure, positioning the company to capture the financial flow as AI workloads move from experimentation into production.

Aug 16, 2026

TechCrunch
Anthropic Claude

TechCrunch

Anthropic CEO Dario Amodei Says AI Backlash Is 'Fundamentally a Crisis of Trust,' Not a Messaging Problem

Dario Amodei pushed back on claims that his cautious risk messaging is fueling public hostility toward AI, arguing in a Fortune interview that the deeper problem is decades of eroding trust in corporations, governments, and the tech industry — not negative framing by AI leaders. He released an essay titled "Policy on the AI Exponential" laying out his views on frontier versus open model regulation, backed by a $200 million Anthropic commitment to research AI's societal impacts.

Aug 16, 2026

One AI workflow, every weekday.

Tutorials, tool reviews, and automation playbooks for solopreneurs running on AI. Short, useful, and free. Unsubscribe anytime.

No pitch. No upsell. One quick AI workflow per weekday.