Short takes on product updates, launches, and pricing changes across the tools we cover.
No hype — just what's relevant if you use these tools professionally.
Google's June 2026 spam update finished rolling out on June 26, 2026, after roughly two days — spanning noon ET June 24 to approximately 2pm ET June 26. Google described it as a "normal" spam update affecting all languages and locations, with no new spam policies introduced, but slightly more significant than the March 2026 spam update. The update targets spam, low-value, and manipulative content — not responsible AI-assisted publishing that produces genuinely useful output. If you use an AI writing tool like OmniSEO, the relevant question is whether your output is genuinely useful and original — not whether AI was involved. The AI blog-writing guide covers what "genuinely useful" means in practice. Sites not violating spam policies should be unaffected; recovery from spam penalties, if triggered, can take months.
Figma Config 2026 (announced June 24) introduced a cluster of changes that push the design-to-code boundary significantly closer. Code Layers brings code directly onto the canvas — clone repos and extract code flows into design layers for testing. Native Animation/Motion lets designers create animations, transitions, and 3D transforms in Figma and export directly to code, removing the rebuild step. Additional announcements include AI-powered shader effects, an enhanced AI assistant that lets users write prompts to create repeatable skills for AI agents and connect external tools (Notion and GitHub confirmed), and AI-generated custom plugins such as layout generators and vector path tracers. A Weavy integration for node-based workflows is coming later in the year. No pricing changes were disclosed. For buyers evaluating AI website builders, the design→code gap Figma is closing is the same gap those tools compete on.
ByteDance unveiled Seedance 2.5 at Volcano Engine's FORCE conference on June 23, 2026 — a video-generation model capable of producing a single native clip up to 30 seconds, including scene changes and tempo shifts, without post-production stitching. The model accepts up to 50 concurrent inputs (reference images, audio, and others), supports post-generation editing while preserving visual style, and handles multi-character film scenes. Seedance 2.5 is currently in enterprise beta; a public launch is targeted for early July 2026. Separately, the existing Seedance 2.0 now offers native 4K output with 10-bit color. For buyers evaluating the AI video tools market, this is a capability signal for where the category is heading — not an immediate buyer recommendation.
Databricks announced Genie One at Data + AI Summit on June 16, 2026 — a generally available agentic AI coworker designed for business teams across marketing, finance, and sales. Genie One automates workflows across structured and unstructured data via a Genie Ontology self-improving context layer, connects to 50+ external apps (Google Drive, Jira, Slack, Confluence, SharePoint), and runs on web, iOS, and Android. Pricing is usage-based, not seat-based: organizations receive $10 free monthly per user with usage-based overages beyond that. Companion launches include Genie Agents (reusable saved workflows), Genie Code (autonomous data-pipeline agent, now GA), Genie App Builder (private preview), and Genie ZeroOps (private preview).
Amazon Bedrock Guardrails announced InvokeGuardrailChecks on June 16, 2026 (AWS Machine Learning Blog) — a new API that lets builders apply content filters, prompt-injection detection, and sensitive-information filters at any point inside an agentic loop without creating, versioning, or managing a guardrail resource. Unlike the existing ApplyGuardrail API, InvokeGuardrailChecks is resourceless: safeguards are specified inline per request with no guardrail ARN required. It operates in detect-only mode and returns numeric severity/confidence scores (discrete values from 0 to 1.0) per safeguard category — so teams set custom thresholds and decide per step whether to block, bypass, retry, or log. IAM permission required: `bedrock:InvokeGuardrailChecks`. The blog walks through setup with a us-east-1 example; regional availability applies per AWS documentation.
Salesforce's Summer '26 release graduated Agentforce multi-agent orchestration from beta to general availability on June 15, 2026, with the rollout beginning June 13. The release ships Atlas Reasoning Engine 3.0 as the coordination layer, lets an orchestrator agent route work to specialist subagents based on their descriptions and actions, adds Agent2Agent (A2A) support for connecting to third-party agents, and brings Salesforce-hosted MCP servers to GA. Cross-platform A2A beyond Salesforce Cloud remains in beta. Salesforce's official release page returned a bot block at the time of writing; the facts here are corroborated across multiple reachable secondary sources.
On June 12, 2026 — just three days after launch — Anthropic disabled access to Claude Fable 5 and Mythos 5 for all users worldwide, complying with a US government export-control directive that ordered the models suspended for any foreign national. Anthropic pulled them for everyone to ensure compliance. Opus 4.8 is the fallback — existing Fable 5 sessions now end in an error, and new sessions route to your default model or Opus 4.8. All of Anthropic's other models are unaffected.
Moonshot AI released Kimi K2.7-Code (model ID kimi-k2.7-code) on June 12, 2026, publishing full weights to Hugging Face under a Modified MIT license. It is a 1-trillion-parameter Mixture-of-Experts model (32B active) tuned for long-horizon agentic coding, priced at just $0.95 / $4.00 per million input/output tokens — a fraction of frontier-tier rates. The catch for operators: every benchmark published so far is Moonshot's own — there are no independent third-party numbers yet. See the best AI coding tools for how it stacks up.
OpenAI and Oracle announced on June 10, 2026 that enterprise customers can apply their existing Oracle cloud commitment — the prepaid Universal Credits companies buy from Oracle — toward OpenAI frontier models and Codex through the OCI Marketplace, with no separate AI procurement required. The move targets Oracle's regulated buyer base — banks, insurers, healthcare, and government — where procurement and compliance teams have often blocked AI adoption. It was enabled by the April 2026 unwinding of OpenAI's Microsoft exclusivity, which freed OpenAI to sell across AWS, Google Cloud, and Oracle.
Two Anthropic deadlines land on June 15, 2026 — four days from now. First, the model IDs claude-sonnet-4-20250514 and claude-opus-4-20250514 are retired: any API request still calling those IDs will fail, with no grace period and no automatic redirect. Second, usage from the Claude Agent SDK and the headless claude -p command moves to a separate credit pool for subscription users, ending the implicit assumption that automation usage is covered by the same limits as interactive chat.
Cursor shipped the biggest Bugbot upgrade since launch on June 10, 2026 — the automated code-review agent is now 3× faster, with 90% of runs finishing under 3 minutes, 22% cheaper per review, and finds 10% more bugs per pass. A new pre-push review command lets developers run Cursor's full review locally before pushing, shifting defect detection earlier in the workflow. Gains are powered by harness improvements and the Composer 2.5 model that now underpins Bugbot.
Ramp's June 2026 AI Index — drawn from aggregated spend data across more than 70,000 US businesses — shows Anthropic reaching roughly 41% business adoption, up from around 34.4% in May, while OpenAI holds at roughly 39.5%. The headline number for budget planning: the most AI-intensive firms (top 1%) spend around $7,449 per employee per month on AI tools, while the median US company on Ramp's platform spends just $11.38. The data covers AI tool spend captured via Ramp's corporate card and bill-pay platform — not a survey, and not whole-market revenue.
Anthropic launched Claude Fable 5 (claude-fable-5) on June 9, 2026 — a new Mythos-class tier above Opus 4.8, priced at $10/$50 per million input/output tokens. Pro, Max, Team, and Enterprise subscribers get Fable 5 included through June 22, 2026; from June 23 onward, usage credits are required. A companion invitation-only model, Claude Mythos 5, released simultaneously under the Project Glasswing program.
OpenAI's Codex app 26.608 (June 9, 2026) ships a guided "Migrate to Codex" onboarding flow that imports six config types from Claude Code and Claude Cowork — including system prompts, custom skills, MCP server configurations, hooks, sub-agent settings, and 30 days of chat history. The same release revamped the plugins screen and expanded Settings search. Companion CLI 0.139.0 adds standalone web search inside code-execution mode, closing a notable capability gap.
Google's Gemini Enterprise release notes confirm that effective June 9, 2026, the feature management toggle that let admins disable Gemini 3.5 Flash is gone. The model is now enabled by default for all users across the Standard, Plus, and Frontline editions and cannot be disabled. The change applies to Global, US, and EU multi-regions, and was pushed back one day from the originally announced June 8 effective date.
At WWDC on June 8, Apple announced Xcode 27 with an autonomous Agent Mode that lets coding agents interact with the iOS simulator, write and run tests, and self-validate fixes. The Foundation Models framework adds image input, a Python SDK, a new Core AI framework for running full local LLMs on Apple Silicon, and an fm CLI for on-device and Private Cloud Compute access from the terminal. Developers with fewer than 2 million App Store downloads can access Apple Foundation Models running on Private Cloud Compute at no cloud API cost, per Apple's press release.
OpenAI shipped Codex CLI 0.138.0 on June 8, 2026, adding a new `/app` command that hands off a live CLI thread into Codex Desktop on macOS and native Windows. The release also exposes saved image file paths to the model for follow-up edits, adds account token-usage visibility for app-server integrations, and expands plugin automation with structured JSON output. Two alpha builds (0.139.0-alpha.1 and alpha.2) followed on June 9. If you're already using Codex CLI for agentic coding, this update meaningfully extends what agents can do without leaving the terminal.
Apple and Google confirmed a multi-year collaboration in January 2026 under which Siri will be rebuilt on Gemini models — and at WWDC on June 8, Apple showed the product for the first time. iOS 27, iPadOS 27, and macOS 27 reportedly introduce a new Extensions system that lets users designate a third-party AI — including Claude, ChatGPT, or Gemini — as their preferred assistant and as the default engine behind Writing Tools and Image Playground, per Bloomberg reporting and WWDC keynote coverage. Apple and Google confirmed the multi-year partnership runs Gemini inside Apple's Private Cloud Compute infrastructure, preserving Apple's privacy model.
Supabase raised $500 million in a Series F round at a $10.5 billion post-money valuation, roughly doubling the valuation it reached in its Series E approximately eight months earlier. The round was led by GIC, Singapore's sovereign wealth fund, with Stripe making a second investment and Salesforce Ventures joining for the first time. The company disclosed that more than 60% of new databases on the platform are now launched by AI coding tools rather than humans, with a 600% year-over-year increase in overall database launches. Anthropic's Claude Code is named as the single largest contributor since the start of 2026.
On June 5, 2026, Google launched the Colab CLI — an open-source, Apache 2.0-licensed command-line tool that lets developers and AI coding agents provision remote Colab GPU and TPU runtimes, execute Python scripts and notebooks, and retrieve artifacts directly from the terminal, without touching a browser. The CLI ships with a bundled COLAB_SKILL.md file so terminal-based AI agents can auto-discover and drive it.
On June 4, 2026, GitHub shipped two new capabilities for GitHub Copilot: one-million-token context windows for working across large codebases and multi-file projects without losing context, and configurable reasoning levels that let developers balance speed against analytical depth — including extended thinking for complex architectural and debugging tasks. Both features launched in VS Code, the Copilot CLI, and the GitHub Copilot App. The catch: choosing a larger context window or higher reasoning level consumes more AI Credits per interaction, which matters for teams already tracking usage under the usage-based billing model that went live June 1.
OpenAI confirmed on June 3, 2026 that it will retire GPT-4.5 and o3 from the ChatGPT model picker in two waves: GPT-4.5 on June 27, 2026 (a 30-day sunset) and o3 on August 26, 2026 (a 90-day sunset). Both changes affect the ChatGPT interface only — o3 remains available via the API, so developers face no disruption. Paid ChatGPT users who manually select either model will need to switch to another model in the picker before the deadlines. OpenAI did not name a specific replacement, though its lineup is consolidating on the GPT-5.5 series.
On June 2, 2026, Cognition officially rebranded Windsurf as Devin Desktop — shifting the product from an AI code editor into a platform for running and managing multiple AI agents at once. The launch introduces the Agent Command Center (a Kanban view across local and cloud agent sessions), Spaces (shared context across sessions, pull requests, and files), the open Agent Client Protocol (ACP) enabling third-party agents including Codex, Claude Agent, and OpenCode to run inside Devin Desktop with access to the same Kanban, Spaces, and shared-context workflow, and Devin Local — a complete Rust rewrite of the Cascade agent with up to 30% greater token efficiency and native subagent support. Existing plans and pricing are unchanged; Cascade legacy support ends July 1, 2026.
At Build 2026 on June 2, 2026, Microsoft announced the infrastructure pieces for Windows as a local AI agent execution platform: new Aion small language models for on-device agentic tasks (including a 14B in-box reasoning model), Microsoft Execution Containers (MXC) for policy-enforced agent sandboxing, and expanded Windows AI runtime APIs. The announcement positions Windows 11 as a local operating system for building, deploying, and governing AI agents — a response to demand for on-device AI that doesn't route sensitive data to the cloud.
At Build 2026 on June 2, 2026, Microsoft launched MAI-Thinking-1, a mid-weight large language model for chat and reasoning, in public preview through Azure AI Foundry. It arrived alongside three other new MAI models (MAI-Image-2.5, MAI-Transcribe-2, MAI-Voice-2) as part of a broader first-party model push — what Microsoft's Build news framed as seven new MAI models overall.
At Microsoft Build 2026 on June 2, GitHub introduced the GitHub Copilot App — a standalone, agent-native desktop experience for running and steering coding agents. It centres on a 'My Work' view across connected repositories, runs each agent session in its own isolated Git worktree so parallel agents don't conflict, and ships alongside the now-generally-available GitHub Copilot SDK. The app is in technical preview for paid Copilot users on Windows, Mac, and Linux.
On June 1, 2026, Salesforce signed a definitive agreement to acquire Contentful — the Berlin-founded headless CMS trusted by over 4,800 brands including IKEA, Vodafone, Electronic Arts, and DoorDash. The acquisition gives Salesforce's Agentforce AI platform a native content layer, enabling AI agents to dynamically assemble and deliver personalised content without manual publishing steps. Financial terms were not disclosed; the deal is expected to close in Q3 of Salesforce's fiscal year 2027.
At Computex 2026 on June 1, Jensen Huang announced Nemotron 3 Ultra — a 550-billion-parameter mixture-of-experts open model designed specifically for agentic coding, research, and enterprise workflows. It ranks first among US open-weights models on Artificial Analysis's Intelligence Index with a score of 48, though it trails China's Kimi K2.6 (54) and closed models like Claude Opus 4.8 (61). Weights ship June 4 on Hugging Face, OpenRouter, and NVIDIA NIM.
GitHub Copilot's transition to usage-based billing — announced April 27 and detailed in our May 20 mechanics explainer — took effect on June 1, 2026. Monthly Pro and Pro+ subscribers migrated automatically. Business and Enterprise organisations received a temporary promotional credit allotment through August that is higher than the standard plan-dollar parity. Here is what the change looks like on the first day it is live, and what to check on your billing dashboard.
NVIDIA launched Cosmos 3 on May 31, 2026 at GTC Taipei — an open frontier omnimodel for physical AI built on a mixture-of-transformers architecture that generates text, images, video, ambient sound, and action trajectories in a single model. It ships with open weights on Hugging Face, is deployable via NVIDIA NIM microservices, and is backed by the Cosmos Coalition — a founding group that includes Black Forest Labs, LTX, and Runway alongside robotics firms.
OpenAI posted multiple robotics engineering job listings on May 31, 2026, reviving a hardware division the company wound down in 2019 after a three-year robotic-hand project. The postings cover electrical, actuator design, simulation environments, and control systems engineering roles paying $210K–$310K base in San Francisco. No product has been announced and no timeline has been given — the hiring push is the signal, not a launch.
On May 29, 2026, xAI made grok-build-0.1 — the dedicated coding model behind its Grok Build CLI — available on the xAI API in public beta, following the May 25 broadening of Grok Build itself to all SuperGrok and X Premium+ subscribers. grok-build-0.1 is a fast agentic coding model with a 256K context window, priced at $1 per million input tokens and $2 per million output (with a $0.20 cached-input rate). Grok Build is xAI's entry into the terminal coding-agent category that Claude Code and Codex already occupy.
ElevenLabs announced Music v2 on May 26, 2026 (updated May 28), a music-generation model that can shift genres mid-track, regenerate individual sections without changing the rest of a song, and embed sound effects directly in a track. It is available now in ElevenMusic and ElevenCreative, with API access coming soon. ElevenLabs cut pricing up to 50% for the API and up to 40% for ElevenCreative self-serve, and says the model is trained only on licensed data and cleared for commercial use.
Anthropic released Claude Opus 4.8 on May 28, 2026, with improved agentic coding accuracy, a cheaper fast mode, and Dynamic Workflows in research preview — a Claude Code feature that runs up to 1,000 parallel subagents on Max, Team, and Enterprise plans to execute codebase-scale tasks autonomously.
YouTube announced on May 27, 2026 that it will automatically apply AI content labels when its systems detect significant photorealistic AI use, even if the creator did not disclose it. Labels for long-form videos move below the player; on Shorts they appear as an overlay. Incorrect labels can be disputed in YouTube Studio — but content made with YouTube's own AI tools carries a permanent disclosure.
Meta launched consumer Plus subscriptions globally on May 27, 2026 — $3.99/mo for Instagram and Facebook, $2.99/mo for WhatsApp — with social features, not AI. Two Meta One AI tiers ($7.99 and $19.99/mo) with advanced reasoning and video/image generation are beginning testing next month in Singapore, Guatemala, and Bolivia only. Free Meta AI access continues unchanged.
Microsoft's May 26, 2026 Copilot Studio update graduates three agent capabilities to general availability — computer-using agents, agent-to-agent communication, and real-time voice — while shipping a redesigned workflows canvas and a new orchestration layer the company says cuts token use by roughly 50%.
Google flipped the default response format for the Gemini Interactions API on May 26 to a new steps-based schema, with the legacy schema set for permanent removal on June 8. Older SDKs and direct REST integrations will break unless they upgrade or pin the temporary opt-out header before the sunset.
Anthropic's May 22 Glasswing update says Claude Mythos Preview found more than 10,000 high- or critical-severity vulnerabilities in its first month, with independent security firms validating 90.6% of the high/critical findings they reviewed. Claude Security — the productized version of the same scanner — is now in public beta for Claude Enterprise customers.
OpenAI's May 22 Codex update graduates Goal Mode from experimental to stable across the app, IDE extensions, and CLI, adds Appshots screen injection on macOS, and lets Codex continue Computer Use after your Mac locks — three changes that push agentic coding further toward unattended, long-horizon runs.
Google launched Gemini Omni Flash at I/O 2026 — a new multimodal video model that takes text, images, audio, and video as input and outputs AI-generated video. It is rolling out free on YouTube Shorts and YouTube Create, and is included on paid Google AI plans starting at $7.99/month.
OpenAI announced on May 19 that it is becoming C2PA-conformant and embedding Google DeepMind's invisible SynthID watermark into all images generated through ChatGPT, Codex, and the OpenAI API — and is previewing a public verification tool anyone can use to check if an image came from an OpenAI model.
Anthropic introduced Routines for scheduled and webhook-triggered Claude Code automations, and Dreams — a research-preview self-improvement loop — at its first European developer conference.
Google introduced Google Pics — an AI image creation and editing tool built on Nano Banana 2, integrating with Workspace and watermarked with SynthID. Limited trusted-tester access today, broader rollout to AI Pro and Ultra subscribers this summer.
Google launched Managed Agents in Public Preview at I/O 2026 — a one-call API that spins up an isolated Linux environment where Gemini can reason, plan, run tools, execute code, and browse the web.
OpenAI co-founder and former Tesla AI lead Andrej Karpathy is joining Anthropic's pre-training team and launching a new effort to use Claude itself to accelerate pretraining research.
All GitHub Copilot plans switched to token-based AI Credits on June 1, 2026. Here's how the new billing model works, what counts toward your credit allotment, and whether agent-mode or Chat-heavy usage will cost more.
Cursor shipped Composer 2.5, its in-house AI coding model, positioned as a frontier-class alternative to Claude Opus 4.7 at a meaningfully lower per-token price.
Alexa+ now produces AI-generated podcast episodes from licensed news sources on user request. For listeners, that's a new feature. For creators, it's a signal worth reading.
OpenAI co-founder Greg Brockman has permanently taken charge of the company's entire product strategy, merging ChatGPT, Codex, and the developer API into a single product organization. The end state is a unified agentic experience — one platform where conversation, coding, and autonomous task execution converge — with a future 'super app' combining Codex, ChatGPT, and the Atlas browser on the roadmap.
Google I/O 2026 opens May 19 at 10am PT. The headline confirmed announcement is Googlebooks — a new category of AI-first Android laptops launching with Acer, ASUS, Dell, HP, and Lenovo this fall. Alongside it: Gemini Intelligence rolling deeper into Android, Chrome, and wearables, and Android XR glasses getting their first public preview.
Upwork's annual In-Demand Skills report found that skills explicitly referencing AI grew 109% year-over-year on the platform. AI video generation leads with 329% growth. AI-enabled freelancers earn approximately 40% more per hour than those not using AI tools. Demand for human expertise in coding, creative work, and marketing has not declined — it has shifted toward AI-fluent specialists.
Mistral has launched its 128B parameter flagship model alongside two major product updates to Le Chat: async cloud coding sessions that run in the background without keeping the interface open, and a Work agentic mode that executes multi-step tasks across tools and files autonomously. The release positions Mistral as the most capable alternative to OpenAI and Anthropic for builders who want frontier-tier AI without full vendor lock-in.
HubSpot released AEO Sensor, a free public dashboard tracking how often ChatGPT, Gemini, and Perplexity cite brands across industries — launching the same week the company disclosed a 27% organic traffic decline for its customer base. No login required. Any operator can benchmark their industry's AI visibility today.
Anthropic is restricting Claude subscribers from running the model through third-party agent harnesses without drawing from a separate credit pool — the first explicit acknowledgment that unlimited AI subscriptions cannot survive autonomous agent workflows. OpenAI immediately responded with 2 free months of Codex for new business customers, directly targeting disaffected Claude agent users.
Runway, the AI video generation startup behind Gen-4.5, added $40 million in annual recurring revenue in Q2 2026 alone — pushing its valuation to $5.3 billion. Co-founder Cristóbal Valenzuela told TechCrunch the company's ambition has shifted: Runway is no longer competing with InVideo and CapCut — it's going after Google's Veo and the Gemini video ecosystem.
Anthropic is negotiating a new funding round of $30 billion to $50 billion at a valuation of up to $950 billion, according to a New York Times report. The round would nearly triple the $380 billion valuation Anthropic closed in February and, if completed, would make Anthropic the most valuable AI lab in the world — passing OpenAI for the first time. Run-rate revenue has reached $14 billion. Claude Code alone is at a $2.5 billion run-rate, more than double its January figure. The round is expected to close by the end of May.
Cursor shipped two material changes in May 2026 that reshape how teams and individual developers use it. On May 8, Cursor went native inside Microsoft Teams — engineers can @mention Cursor in any channel to delegate tasks to cloud agents and have them open PRs for review. Separately, Bugbot is moving from per-seat to usage-based billing for both Teams and Individual plans, effective with the next renewal after June 8. The combination removes the two biggest friction points solo devs and small teams cited about Cursor's review tooling.
Anthropic officially launched Claude for Small Business on May 14, packaging 15 pre-built workflows, native connectors for QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365, and a 10-city free training tour. Simultaneously, Anthropic and the Bill & Melinda Gates Foundation announced a $200 million partnership to bring AI business tools to underserved entrepreneurs — the clearest signal yet that Anthropic sees SMBs as a primary growth market, not a downstream afterthought.
In a 12-day window ending May 13, four AI labs — Z.ai, MiniMax, Moonshot, and DeepSeek — shipped open-weights coding models that match or closely approach frontier performance on standard benchmarks. All four are downloadable and self-hostable. The compressed release timeline and open-weights distribution mark a significant shift in the competitive landscape for AI coding tools: capable coding models are now available outside closed API ecosystems.
Baiju Bhatt, co-founder of Robinhood, raised $275 million at a $2 billion valuation for Cowboy Space Corporation, which plans to build rockets whose upper stages double as AI compute nodes in low Earth orbit. Each satellite carries 800 GPUs powered by solar energy. The company's constellation is called Stampede and aims to launch before the end of 2028.
Anthropic notified Claude Team and Enterprise admins this week that three features — legacy Artifacts, custom Styles, and Project file upload search — will be retired by June 15. All three are being replaced by VM-based equivalents that require code execution and file creation to be enabled. It is not just a deprecation notice: it is a signal that Anthropic is consolidating all of Claude's power features onto a single sandboxed execution architecture.
OpenAI launched the OpenAI Deployment Company on May 12 with $4 billion in backing from TPG, McKinsey, Bain Capital, and 16 other firms. It acquires AI consulting firms and places deployment engineers directly inside enterprises — a direct move into the market currently owned by Accenture, Deloitte, and McKinsey itself.
At the Android Show on May 12 — one week before Google I/O — Google announced Gemini Intelligence, which embeds Gemini directly into Android as an ambient AI layer rather than a standalone app. It also unveiled Googlebook, a new laptop line built around Gemini launching fall 2026 with Acer, Dell, HP, and Lenovo.
Cactus Compute released Needle on May 12 — a 26 million parameter model distilled from Gemini 3.1 that runs on consumer hardware at 6,000 tokens/second. MIT licensed, open source, and built specifically for tool calling. It is the clearest signal yet that capable AI agents are moving off the cloud and onto the device.
OpenAI replaced GPT-5.3 with GPT-5.5 Instant as the default ChatGPT model on May 5, citing a 52.5% reduction in hallucinated claims across medicine, law, and finance — plus new cross-session memory and a native integration for Excel and Google Sheets.
Anthropic launched ten ready-to-run Claude agent templates for financial services at an invite-only briefing in New York on May 5, covering KYC screening, pitchbook generation, earnings review, financial modeling, and month-end close automation — alongside an announcement of Claude's integration with Microsoft 365.
Perplexity released its Personal Computer desktop app for all Mac users on May 7, giving AI agents direct access to local files, installed applications, and third-party connectors for multi-step workflows — requiring a Pro or Max subscription.
Microsoft's Q1 2026 AI diffusion report shows 17.8% of the world's working-age population now uses generative AI — up from 16.3% in late 2025 — but the gap between rich and developing nations is accelerating, not closing.
Moonshot AI, maker of the Kimi chatbot, closed a $2 billion funding round led by Meituan on May 7, pushing its valuation past $20 billion and cementing Kimi K2.6 as the second-most-used large language model on OpenRouter.
EU negotiators reached a provisional deal on May 7 to streamline the AI Act, pushing high-risk obligations from August 2026 to December 2 2027 while tightening the deadline for labeling AI-generated content from six months to three.
Anthropic signed a deal with SpaceX to access 300 megawatts of compute across 220,000 Nvidia GPUs at the Colossus 1 data center in Memphis, with the company saying this will directly raise usage limits for Claude Pro and Max subscribers.
InVideo AI expanded its text-to-video capabilities with 50 new AI avatar voice profiles and a custom brand kit feature, letting Plus subscribers lock in brand colors, fonts and preferred avatars across all projects.
Anthropic partnered with Blackstone, Hellman and Friedman, and Goldman Sachs to launch a new $1.5B enterprise AI services firm that embeds Claude — with Anthropic engineers on-site — directly into mid-market companies owned by the financial partners.
Notion moved Custom Agents to a paid credit model starting May 4, restricting creation to Business and Enterprise plan users, with admins now able to set per-agent and workspace-level credit limits.
Notion merged its Calendar product with Notion AI, enabling automatic deadline extraction from project databases, smart scheduling suggestions and daily briefings generated from your workspace data.
Bret Taylor's AI customer-service agent startup Sierra raised $950M at a $15.8B post-money valuation in a round led by Tiger Global and GV, with ARR jumping from $100M in November to $150M by February.
Midjourney released v8.1 across its web app and Discord on April 30, delivering native 2048x2048 HD output without a separate upscale step and standard job rendering 4-5x faster than earlier versions.
Zapier released two features in open beta -- Tables AI Enrich, which auto-populates multiple structured fields across records from a single prompt, and Agent Versioning, which separates draft and published agent states so live automations are not disrupted by edits.
Anthropic released Claude Opus 4.7 as generally available on April 16, with 3.3x improved vision supporting images up to 2,576px, software engineering gains over Opus 4.6, a 1M context window, and no long-context pricing premium.
Grammarly introduced a full-document rewrite feature for Premium and Business subscribers, enabling AI-powered restructuring and tone adjustment at the document level -- not just sentence by sentence.
Anthropic added six enterprise features to Claude for Teams on April 9 2026 -- role-based access controls, group spend limits, usage analytics, expanded audit logging, a Zoom MCP connector and per-tool permission controls -- alongside Managed Agents in public beta for teams building long-running AI workflows.
Anthropic unveiled Project Glasswing, a defensive cybersecurity coalition built around the unreleased Claude Mythos Preview model -- which has already found thousands of zero-day vulnerabilities, including a 27-year-old SACK bug in OpenBSD.
Black Forest Labs's Flux 2 has quietly become the benchmark leader for open-source image generation in 2026, outperforming Stability AI's SD 3.5 on photorealism, prompt adherence and native 2K output.
Zapier launched AI Guardrails on March 30 2026 -- a built-in safety layer for its Agents feature that checks outputs and routes ambiguous actions for human review before executing across its 7,000+ app integrations.
Suno released v5.5 with three major additions: Voices (voice cloning for Pro/Premier subscribers), Custom Models trained on your own catalog, and My Taste personalisation that learns your preferences over time.
Hostinger upgraded its Horizons AI Website Builder with one-click section regeneration, letting users rewrite any page block with a new prompt without starting from scratch.
Microsoft rolled out deep Copilot integration across Teams and Outlook in March 2026, enabling meeting summaries, draft replies, calendar conflict resolution and action-item extraction directly inside both apps.
Anthropic launched Claude computer use on March 23-24 2026, giving Pro and Max subscribers the ability to delegate full desktop tasks to Claude -- letting the model open apps, navigate browsers, read screens and complete multi-step work autonomously.
Adobe expanded Firefly's video and image creation capabilities with AI-powered custom models, letting Creative Cloud subscribers train Firefly on their own brand assets for consistent output across campaigns.
Coursera and Google launched the AI for Business Professional Certificate on February 19 2026, covering prompt engineering, AI strategy and workflow automation -- designed to be completable in under three months.
ElevenLabs released Eleven v3 into general availability in February 2026, bringing 74-language support and Audio Tags -- bracketed emotion instructions embedded directly in scripts that control tone, pacing and delivery without re-recording.
Perplexity launched Model Council on February 5 2026 -- available to Max and Enterprise Max subscribers -- running a single search query through multiple frontier AI models simultaneously and returning parallel answers for side-by-side comparison.
Runway launched Gen-4.5 on December 1 2025, making it the top-rated commercial video model available across all paid plans, while its GWM-1 General World Model -- capable of simulating explorable environments, conversational avatars and robotic manipulation -- shipped shortly after.
Workflow automation platform n8n closed a $180M Series C round on October 9 2025, led by Accel and Meritech with participation from Deutsche Telekom T.Capital, valuing the company at $2.5 billion.
OpenAI released GPT-5 on August 7 2025, making its most capable model available to all ChatGPT users -- including free-tier accounts -- with native multimodal reasoning across text, images, audio and structured data.
NordVPN rolled out post-quantum cryptography across all its applications on May 22 2025 -- covering Windows, macOS, iOS, Android, Android TV and tvOS -- protecting connections against future quantum-computing decryption attacks.
Notion released Research Mode in May 2025, letting users run natural-language queries across linked databases, wikis and documents -- surfacing answers with inline citations drawn from their own workspace.
Figma announced Figma Make at Config 2025 on May 7 2025 -- an AI prototyping system that converts written instructions into fully interactive, structured design files with working logic and component hierarchy built in, ready for developer handoff.
Google DeepMind released Gemini 2.5 Pro in experimental preview on March 25 2025, posting the highest score on LMArena and top results on GPQA Diamond and AIME 2025 -- making it the strongest reasoning model Google had shipped to date.