Market intelligence

The market in one breath.

Weekly digests from the WWT radar, weekday briefings from the automated intel pipeline, and the industry timeline behind the scores — the raw feed, drafts included. For the curated, editor-published digest see Last Week in AI →

Industry timeline

Weekly radar sync + weekday daily intel, oldest → newest

Scroll with your mouse or trackpad, drag the canvas, or use the arrow keys to move along the timeline.

214 events

Significance: High (5) Notable (4) Standard (3)

Dec 2023

Jan 2024

Feb 2024

Mar 2024

Apr 2024

Jun 2024

Aug 2024

Sep 2024

Oct 2024

Nov 2024

Dec 2024

Jan 2025

Feb 2025

Apr 2025

May 2025

Jun 2025

Jul 2025

Aug 2025

Oct 2025

Nov 2025

Dec 2025

Jan 2026

Feb 2026

Mar 2026

Apr 2026

May 2026

Jun 2026

Jul 2026

Model releaseDec 6

Google launches Gemini 1.0

Google DeepMind launches Gemini 1.0 in three sizes (Ultra, Pro, Nano), its first natively multimodal model competing with GPT-4 across coding and reasoning benchmarks.

Weekly radargemini-code-assist

Policy / ResearchDec 8

EU AI Act reaches political agreement

EU Parliament and Council reach political agreement on the AI Act, the world's first comprehensive horizontal AI regulation with risk-based classification for AI systems.

Weekly radar

Open sourceDec 11

Mistral releases Mixtral 8x7B open-weight

Mistral AI releases Mixtral 8x7B, the first open-weight mixture-of-experts model, outperforming Llama 2 70B on most benchmarks with 6x faster inference.

Weekly radar

Pricing / PlatformJan 10

OpenAI launches ChatGPT Team tier at $25/seat

OpenAI launches ChatGPT Team at $25/seat/month with GPT-4, DALL-E 3, and data privacy guarantees, creating a mid-market tier for small engineering teams.

Weekly radar

Open sourceJan 24

Zed goes open source under GPL/AGPL licenses

Zed releases its high-performance code editor as open source — GPL for the editor, AGPL for server components, Apache 2.0 for the GPUI framework.

Weekly radar

Open sourceJan 29

Meta releases Code Llama 70B

Meta releases Code Llama 70B, its largest open-weight coding model, approaching GPT-3.5 performance on code generation benchmarks.

Weekly radar

Model releaseFeb 15

Google announces Gemini 1.5 Pro with 1M-token context

Google announces Gemini 1.5 Pro with a 1M-token context window, enabling full-codebase understanding in a single prompt and redefining context limits for coding models.

Weekly radargemini-code-assist

Open sourceFeb 21

Google open-sources Gemma

Google releases Gemma 2B and 7B, lightweight open-source models built from Gemini research, enabling local AI development on consumer hardware.

Weekly radar

Model releaseMar 4

Anthropic releases Claude 3 family

Anthropic ships Claude 3 Haiku, Sonnet, and Opus — establishing three performance tiers with Opus matching GPT-4 and setting new state-of-the-art coding benchmarks.

Weekly radarclaude-code

Product launchMar 12

Devin announced as first AI software engineer

Cognition demonstrates Devin, an autonomous AI software engineer capable of planning, coding, and debugging — sparking the "agentic coding" category. GA followed in December 2024.

Weekly radardevin

Open sourceMar 12

OpenDevin launches as open-source AI coding agent

Binyuan Hui and Junyang Lin launch OpenDevin (later OpenHands) on GitHub, quickly becoming #1 trending. MIT-licensed platform for autonomous software development agents.

Weekly radar

Open sourceApr 18

Meta releases Llama 3 weights

Meta open-sources Llama 3 8B and 70B with a permissive license, triggering a wave of self-hosted coding assistants and fine-tuned models.

Weekly radarcontinue

Product launchJun 17

Warp launches Agent Mode for multi-step AI workflows

Warp ships Agent Mode, enabling AI to execute commands, generate code diffs, and self-correct errors in the terminal via natural language.

Weekly radar

Policy / ResearchAug 13

SWE-bench Verified benchmark published

Princeton publishes SWE-bench Verified, a human-validated subset of SWE-bench that becomes the standard for evaluating agentic coding tools.

Weekly radar

Product launchAug 20

Zed AI launches with Claude 3.5 Sonnet integration

Zed introduces AI-powered assistant panel and inline code transformations powered by Anthropic's Claude 3.5 Sonnet, including Fast Edit Mode for near-real-time refactoring.

Weekly radar

Product launchSep 26

Aider introduces Architect/Editor dual-model approach

Aider ships Architect mode, separating code reasoning from editing using two LLMs. Achieves 85% SOTA with o1-preview as Architect and DeepSeek as Editor.

Weekly radar

Product launchOct 29

GitHub Universe: multi-model Copilot and Spark

At GitHub Universe 2024, GitHub opens Copilot to Claude 3.5 Sonnet, Gemini 1.5 Pro, and o1 models, ending GPT exclusivity. Also launches Spark, an AI-native app builder.

Weekly radargithub-copilot

Business moveNov 22

Amazon invests additional $4B in Anthropic

Amazon invests another $4B in Anthropic, bringing total to $8B. Anthropic names AWS as primary training partner and commits to Trainium and Inferentia chips.

Weekly radarclaude-code

Product launchDec 10

Replit Agent exits beta

Replit Agent moves to general availability, targeting citizen developers and rapid prototyping with autonomous app building capabilities.

Weekly radarreplit

Product launchDec 11

Google announces Jules AI coding agent with Gemini 2.0

Google Labs unveils Jules, an experimental asynchronous AI coding agent powered by Gemini 2.0, operating within GitHub workflows for bug fixes and multi-file edits.

Weekly radar

Pricing / PlatformDec 18

GitHub Copilot Free tier launches with 2K completions/mo

GitHub launches a free tier of Copilot with 2,000 completions and 50 chat messages per month, commoditizing basic AI code completion.

Weekly radargithub-copilot

Open sourceJan 30

Mistral releases Mistral Small 3 under Apache 2.0

Mistral releases Mistral Small 3, a 24B parameter Apache 2.0 model competitive with Llama 3.3 70B and Qwen 32B, enabling efficient self-hosted AI coding workflows.

Weekly radar

Policy / ResearchFeb 2

Andrej Karpathy coins "vibe coding" term

Former OpenAI co-founder Andrej Karpathy coins "vibe coding" to describe AI-assisted development where developers accept LLM-generated code without close review.

Weekly radarcursor

Pricing / PlatformApr 2

Tabnine sunsets free Basic tier, shifts to enterprise

Tabnine discontinues its free Basic tier, transitioning users to a limited Dev Preview. Marks a broader industry shift from free-tier growth to enterprise-focused monetization.

Weekly radartabnine

Pricing / PlatformApr 3

Devin 2.0 launches at $20/mo, down from $500

Cognition AI launches Devin 2.0 with agent-native IDE, parallel sessions, and 96% price cut from $500 to $20/mo starting plan.

Weekly radardevin

Open sourceApr 16

OpenAI open-sources Codex CLI under Apache 2.0

OpenAI publishes Codex CLI to GitHub as open-source terminal coding agent built in TypeScript. Offers $1M in API grants to developers.

Weekly radarcodex

Open sourceApr 28

Alibaba releases Qwen 3 open-source model family

Alibaba releases Qwen 3 family under Apache 2.0 with 8 model sizes including 235B MoE, trained on 36T tokens with hybrid thinking mode for code generation and reasoning.

Weekly radaropenhands

Product launchMay 7

Zed introduces Agentic Editing

Zed launches agentic AI coding where the AI writes code and runs tests autonomously, with streaming edits and multibuffer review capabilities.

Weekly radarzed

Product launchMay 20

Jules enters public beta at Google I/O 2025

Google launches Jules into global public beta at I/O, offering 5 free daily tasks with Gemini 2.5 Pro, GitHub integration, and secure cloud sandboxes.

Weekly radarjules

Product launchJun 25

Google launches Gemini CLI as open-source terminal agent

Google releases Gemini CLI under Apache 2.0 with Gemini 2.5 Pro, 60 requests/min free tier, and Model Context Protocol support. Reaches 1M users within 3 months.

Weekly radargemini-cli

Product launchJun 25

Warp 2.0 launches as first Agentic Development Environment

Warp ships 2.0, integrating Code, Agents, Terminal, and Drive into a unified ADE. Scores #1 on Terminal-Bench (52%) and 71% on SWE-bench Verified.

Weekly radarwarp

ShutdownJun 25

Sourcegraph sunsets Cody Free and Pro tiers

Sourcegraph discontinues Cody Free, Pro, and Enterprise Starter plans, pivoting to Amp for agentic workflows. Cody Enterprise remains supported.

Weekly radarcody

Product launchJul 14

Kiro launches in preview from AWS

AWS introduces Kiro in preview at AWS Summit NYC, a spec-driven development IDE that generates implementation from structured requirements and hooks into AWS services. GA follows in November 2025.

Weekly radarkiro

Business moveJul 14

Windsurf carved up by Google and Cognition after OpenAI deal collapses

OpenAI's $3B acquisition of Windsurf collapses over Microsoft IP rights. Google executes a $2.4B reverse acquihire of CEO and engineers; Cognition acquires remaining assets, IP, and ~210 employees within 72 hours.

Weekly radarwindsurf

Policy / ResearchJul 17

Replit incident exposes agent control gaps

A widely-reported incident shows Replit Agent executing unintended destructive commands, highlighting governance gaps in autonomous coding tools.

Weekly radarreplit

Policy / ResearchAug 2

EU AI Act enforcement begins

EU AI Act general-purpose AI provisions take effect, requiring transparency obligations for foundation model providers used in developer tools.

Weekly radar

Open sourceAug 5

OpenAI releases gpt-oss open-weight models

OpenAI releases gpt-oss-120b and gpt-oss-20b, its first open-weight models since GPT-2, under Apache 2.0 — enabling self-hosted AI workflows at low cost.

Weekly radar

Product launchAug 6

Jules exits beta with tiered pricing and GA launch

Google launches Jules to general availability with 140K+ beta code improvements, $19.99 Pro and $124.99 Ultra plans, and multimodal support.

Weekly radarjules

Product launchAug 11

Vercel launches

Weekly radarv0

Product launchAug 13

Figma Make (formerly AI) reaches GA

Figma Make ships as a general-availability feature, generating production React/HTML from Figma designs with component mapping.

Weekly radarfigma-make

Product launchOct 8

Gemini CLI launches extensions framework with Figma, Stripe

Google introduces third-party extension framework for Gemini CLI with launch partners Figma, Stripe, Dynatrace, and Elastic via Model Context Protocol.

Weekly radargemini-cli

Product launchNov 5

Tabnine launches Agentic platform with autonomous coding agents

Tabnine ships its Agentic platform with AI agents that complete entire workflows autonomously, powered by its Enterprise Context Engine for personalized assistance.

Weekly radartabnine

Business moveNov 13

Anysphere (Cursor) raises $2.3B Series D at $29.3B

Cursor's parent Anysphere closes $2.3B Series D co-led by Accel and Coatue at $29.3B valuation, with Google, Nvidia, and a16z participating. Annualized revenue exceeds $1B.

Weekly radarcursor

Product launchDec 9

Agentic Tools Radar launches

First public release of the Agentic Developer Tools Radar, tracking 40+ tools across 5 dimensions with systematic evaluation methodology.

Weekly radar

Open sourceDec 9

MCP spec 2025-11-25 released, donated to Linux Foundation

Anthropic releases MCP spec version 2025-11-25 and donates the protocol to the Agentic AI Foundation (Linux Foundation). Cursor, Windsurf, Claude Code, and VS Code all ship MCP client support.

Weekly radarclaude-code-cursor-windsurf

ShutdownDec 22

JetBrains discontinues Fleet IDE

JetBrains announces discontinuation of Fleet, its lightweight cloud IDE, to focus investment on IntelliJ-based IDEs and AI Assistant integration.

Weekly radarjetbrains-ai-assistant

Open sourceJan 1

Cline named fastest growing AI project in GitHub Octoverse 2025

GitHub Octoverse 2025 recognizes Cline as the fastest growing AI open-source project with 4,704% year-over-year contributor growth.

Weekly radarcline

Policy / ResearchJan 1

California SB 53 frontier AI transparency law takes effect

California SB 53 takes effect, requiring frontier model developers to publish transparency reports, safety frameworks, and report critical incidents within 15 days.

Weekly radarcodex

Product launchJan 7

Claude Code 2.1.0 ships skills and session teleportation

Claude Code 2.1.0 ships with skills hot-reload, /teleport session handoff to

Weekly radarclaude-code

Product launchJan 12

Anthropic launches Claude Cowork desktop agent

Anthropic launches Cowork, a desktop agent giving Claude file access for organizing, creating, and editing documents without coding. Released as macOS research preview.

Weekly radarclaude-code

Product launchJan 15

Replit Agent 3 adds mobile app building

Replit Agent 3 gains ability to build and publish complete iOS and Android apps from natural language prompts.

Weekly radarreplit

Product launchJan 21

Cognition launches Devin Review

Cognition launches Devin Review, a free AI code review tool that analyzes PRs, groups related changes, and detects bugs on any GitHub PR.

Weekly radardevin

Product launchJan 22

Cursor 2.4 ships subagents and skills

Cursor 2.4 adds subagents, skills, Plan/Ask modes in CLI, cloud agent handoffs, and 10-20x faster hooks.

Weekly radarcursor

Product launchJan 22

JetBrains integrates OpenAI Codex across all IDEs

JetBrains adds native OpenAI Codex support across all JetBrains IDEs with free API credits for early adopters.

Weekly radarjetbrains-ai-assistant

Open sourceJan 27

Moonshot releases Kimi K2.5 open-source

Moonshot AI releases Kimi K2.5, a 1T-parameter open-source MoE model with Agent Swarm architecture supporting up to 100 parallel sub-agents.

Weekly radaropenhands

Pricing / PlatformJan 27

Google AI Plus expands to 35 markets at $7.99/mo

Google expands AI Plus to 35 countries at $7.99/mo with Gemini 3 Pro access, undercutting $20/mo competitors.

Weekly radargemini-code-assist

Product launchJan 30

Windsurf ships Arena Mode and Plan Mode

Windsurf ships Arena Mode for side-by-side model comparison in-IDE, plus Plan Mode for structured task planning workflows.

Weekly radarwindsurf

Product launchFeb 2

OpenAI launches Codex macOS app

OpenAI launches a native Codex macOS app for managing parallel AI coding agents with Git worktree isolation and 30-minute autonomous sessions.

Weekly radarcodex

Product launchFeb 3

Apple Xcode 26.3 ships agentic coding with Claude and Codex

Apple ships Xcode 26.3 with built-in agentic coding — integrating Claude Agent and OpenAI Codex for autonomous task execution via MCP.

Weekly radarclaude-code-codex

Product launchFeb 3

Snowflake launches Cortex Code AI coding agent

Snowflake launches Cortex Code, a data-native AI coding agent for automating data engineering and analytics development, with CLI now GA.

Weekly radarcodex

Policy / ResearchFeb 3

International AI Safety Report 2026 published

The second International AI Safety Report publishes, authored by 100+ experts across 30+ countries, covering frontier AI risks and safety governance.

Weekly radarcodex

Product launchFeb 4

GitHub opens Agent HQ to Claude and Codex

GitHub opens Agent HQ to Claude and OpenAI Codex alongside Copilot, enabling multi-agent coding in GitHub, GitHub Mobile, and VS Code for Pro+ and Enterprise users.

Weekly radargithub-copilot-claude-code-codex

Product launchFeb 4

Qodo 2.0 launches with multi-agent code review system

Qodo releases second-gen platform with multi-agent system and advanced context engineering, achieving 11% better precision on AI code review.

Weekly radarqodo

Model releaseFeb 5

Anthropic releases Claude Opus 4.6

Anthropic ships Claude Opus 4.6 with 1M-token context, agent teams for multi-agent collaboration, and state-of-the-art coding and reasoning performance.

Weekly radarclaude-code

Model releaseFeb 5

OpenAI ships GPT-5.3-Codex

OpenAI ships GPT-5.3-Codex 20 minutes after Anthropic's Opus 4.6 — its fastest and most capable agentic coding model with a first-ever High cybersecurity rating.

Weekly radarcodex

Product launchFeb 5

OpenAI launches Frontier enterprise agent platform

OpenAI launches Frontier, an enterprise platform for building and managing AI agents across business operations. Initial customers include Uber, State Farm, and Intuit.

Weekly radarcodex

Product launchFeb 10

Visual Studio 2026 ships with AI Profiler Agent

Microsoft releases Visual Studio 2026 with an AI Profiler Agent that uses GitHub Copilot to diagnose performance issues without profiling expertise.

Weekly radargithub-copilot

Product launchFeb 13

Cline CLI 2.0 brings agents to the terminal

Cline ships CLI 2.0 with interactive and autonomous terminal modes, parallel agent processing, and ACP support for Zed, Neovim, and Emacs.

Weekly radarcline

Product launchFeb 13

OpenAI ships Lockdown Mode for ChatGPT and Codex

OpenAI introduces Lockdown Mode restricting data sharing and external integrations for enterprise ChatGPT and Codex deployments.

Weekly radarcodex

Open sourceFeb 16

Alibaba releases Qwen 3.5 open-weight under Apache 2.0

Alibaba releases Qwen 3.5 (397B MoE, 17B active) under Apache 2.0 with 201-language support, claiming benchmark parity with GPT-5.2 and Claude Opus 4.5.

Weekly radaropenhands

Product launchFeb 17

GitHub Agentic Workflows enters technical preview

GitHub launches Agentic Workflows in technical preview, enabling Copilot agents to autonomously create PRs, run tests, and fix CI failures.

Weekly radargithub-copilot

Policy / ResearchFeb 17

Amazon mandates weekly Kiro usage, engineers push back

Amazon requires engineers to use Kiro weekly, sparking internal debate about mandatory AI tool adoption policies.

Weekly radarkiro

Business moveMar 4

Goodfire Raises $50M Series A to Advance AI Interpretability Research

Goodfire, an AI interpretability startup, announced a $50M Series A funding round to advance its research into understanding how AI models work internally. The company focuses on interpretability infrastructure, including work on harvesting activations from trillion-parameter models and using interp

Weekly radar

Policy / ResearchMar 4

MIT Study Highlights Governance Gaps in Autonomous AI Agents Used by Law Firms

An article published on Above the Law examines the risks of deploying autonomous AI agents in law firm environments, drawing on a recent MIT-led study that evaluated dozens of widely used AI agent systems. The study found that many of these tools lack basic monitoring, meaningful transparency, and r

Weekly radar

Product launchMar 5

Stargate AI Data Center Initiative Delayed by Partner Disputes Over Control and Ownership

The Stargate AI data center initiative — a $500 billion joint venture between OpenAI, Oracle, and SoftBank announced by President Trump in January 2025 — has reportedly been delayed due to disagreements among the partners over who would have ultimate control of the planned facilities. Oracle had sec

Weekly radar

Business moveMar 5

Cognition Acquires Windsurf, Combining AI Coding Agent and IDE Capabilities

Cognition, the company behind the AI coding agent Devin, has acquired Windsurf, the AI-powered IDE formerly known as Codeium. The acquisition comes days after Google DeepMind hired Windsurf's CEO and research leaders, and months after OpenAI had offered to buy the company. The deal gives Cognition a

Weekly radar

Product launchMar 6

Stargate AI Infrastructure Project Stalls Over Partner Disputes on Roles and Funding

The Stargate project, a $500 billion mega-scale AI infrastructure initiative announced by President Trump in January 2025, has reportedly been adrift for over a year without moving into full-scale operation. The project, designed to build massive data centers across the United States to secure U.S.

Weekly radar

Business moveMar 6

Goodfire Reaches $1.25B Valuation as AI Interpretability Startup Pivots Toward Model Training Integration

Goodfire, an AI interpretability startup, has reached a $1.25 billion valuation as of February 2026. The company develops mechanistic interpretability tools, including its Ember API, designed to make neural network internals more transparent and steerable. The startup's core work sits at the interse

Weekly radar

Model releaseMar 25

Overnight: Open-Source Tool Uses Claude Code History to Predict and Execute Developer Messages During Downtime

A new open-source project called 'Overnight' (GitHub: yail259/overnight) proposes an approach to autonomous coding that builds a behavioral profile from a developer's Claude Code interaction history and then predicts what the developer would type next. Rather than running batch automation scripts, i

Weekly radar

Model releaseMar 25

H Company Releases Holotron-12B Open Model for High-Throughput Computer Use Agents

H Company has published a blog post on Hugging Face introducing Holotron-12B, a 12-billion parameter model designed specifically for computer use agent tasks with an emphasis on high throughput. The model appears to be positioned as an open-weight solution for building agents that can autonomously i

Weekly radar

Model releaseMar 25

Google's New Gemini Pro Model Achieves Record Benchmark Scores

Google has released a new version of its Gemini Pro model that reportedly sets new records across multiple AI benchmarks. While the full article content was not accessible due to the Google News redirect, the headline indicates continued competitive pressure in the foundation model space, with Googl

Weekly radar

Model releaseMar 25

Google Releases Gemini 3.1 Pro with Updated Benchmark Performance

Google has released Gemini 3.1 Pro, the latest iteration of its flagship large language model family. While the full article content was not accessible due to a Google News redirect, the title indicates the release includes benchmark performance data and availability information for users looking to

Weekly radar

Model releaseMar 25

Ginkgo Bioworks Autonomous Laboratory Powered by GPT-5 Achieves 40% Improvement Over Scientific Benchmark

Ginkgo Bioworks has announced that its autonomous laboratory system, driven by OpenAI's GPT-5, has achieved a 40% improvement over an unspecified state-of-the-art scientific benchmark. The system reportedly integrates GPT-5's reasoning and planning capabilities to autonomously design, execute, and i

Weekly radar

Model releaseMar 25

DeepSeek Featured in Britannica Encyclopedia Entry

Britannica has published an encyclopedia entry on DeepSeek, the Chinese AI company that gained significant attention for its open-weight large language models. The inclusion of DeepSeek in Britannica signals the company's growing mainstream recognition and its establishment as a notable entity in th

Weekly radar

Model releaseMar 25

Cursor Releases Composer 2 Coding Model, Outperforms Claude Opus 4.6 but Trails GPT-5.4

Cursor has released Composer 2, a new coding-specific model integrated into its AI-powered code editor. According to the report, Composer 2 surpasses Anthropic's Claude Opus 4.6 on coding benchmarks but still falls behind OpenAI's GPT-5.4, illustrating the intensifying competition in AI-assisted dev

Weekly radar

Model releaseMar 25

Anthropic Releases Claude Opus 4.6 with 1M-Token Context Window and Enhanced Scientific Reasoning

Anthropic has released Claude Opus 4.6, a new model variant specifically targeting research workflows. The model features a 1 million-token context window and improved scientific reasoning capabilities, representing a significant expansion of Claude's capacity for processing large documents, dataset

Weekly radar

Model releaseMar 25

Qwen 3.5, GLM 5, and MiniMax 2.5: Chinese AI Labs Push the Frontier with Latest Open Model Releases

Chinese AI labs have released a new wave of frontier-class open models, including Alibaba's Qwen 3.5, Zhipu AI's GLM 5, and MiniMax 2.5. This batch of releases, cataloged as part of the 'open artifacts' tracking series by Interconnects AI, represents a significant push by Chinese organizations to co

Weekly radar

Model releaseMar 25

China's Open-Source AI Strategy Builds Self-Reinforcing Advantage Against ChatGPT

China's open-source AI ecosystem is reportedly building a self-reinforcing competitive edge against OpenAI's ChatGPT and other closed-source Western models. The strategy leverages open-source releases from Chinese AI labs to accelerate community-driven development, talent acquisition, and ecosystem

Weekly radar

Model releaseMar 25

Anthropic Introduces Claude Sonnet 4.6

Anthropic has announced Claude Sonnet 4.6, the latest iteration in its Sonnet model line. While the full article content was not accessible due to a Google News redirect, the title confirms a new model release from Anthropic, one of the leading foundation model providers and a key partner in WWT's A

Weekly radar

Model releaseMar 25

Anthropic Reaches General Availability of 1M Context Windows, Following Gemini and OpenAI

Latent.Space's AINews roundup covers a relatively quiet news day, using the lull to reflect on Anthropic's general availability release of 1 million token context windows for Claude. This milestone places Anthropic behind both Google's Gemini and OpenAI, which had already shipped comparable or large

Weekly radar

Product launchMar 25

OpenAI Shuts Down Sora Standalone App

OpenAI has reportedly shut down the Sora app, its standalone consumer-facing application for AI video generation. The article, sourced via Google News from StartupHub.ai, provides limited detail due to content truncation, but the headline indicates the discontinuation of the dedicated Sora applicati

Weekly radar

Product launchMar 25

NXP and Hugging Face Demonstrate Vision-Language-Action Model Fine-Tuning and Deployment on Embedded Robotics Platforms

NXP Semiconductors published a technical blog post on Hugging Face detailing their workflow for bringing Vision-Language-Action (VLA) models to embedded robotics platforms. The post covers the full pipeline from dataset recording for robotic manipulation tasks, fine-tuning VLA models (likely based o

Weekly radar

Product launchMar 25

NVIDIA GTC 2026: OpenClaw Robotics Platform, Vera CPU Launch, and $1T Sales Backlog Announced

At NVIDIA's GTC 2026, CEO Jensen Huang made several major announcements including the launch of OpenClaw, an open robotics platform, and the Vera CPU, which marks NVIDIA's entry into the CPU market to complement its GPU ecosystem. Perhaps most notably, Huang disclosed that NVIDIA has accumulated a $

Weekly radar

Product launchMar 25

JetBrains Launches Qure: Desktop App That Converts Manual QA Flows into E2E Test Code

JetBrains has released Qure, a desktop application described as an "AI copilot for web automation" that records manual web testing flows and generates end-to-end test code tailored to a user's existing repository conventions. The product was shared on Hacker News as a Show HN post, pointing to quret

Weekly radar

Product launchMar 25

FedEx Rolls Out AI Training Program for Over 400,000 Employees to Build Promotion-Ready Skills

FedEx has launched a large-scale AI training initiative targeting its workforce of over 400,000 employees, framing it as 'promotion-ready' AI education. The program represents one of the largest known enterprise AI upskilling efforts by a non-tech company, signaling that AI literacy and capability d

Weekly radar

Product launchMar 25

Apple Xcode 26.3 Adds Model Context Protocol Support for AI-Assisted Coding

Apple has released Xcode 26.3 with integrated support for the Model Context Protocol (MCP), enabling what the industry has termed 'vibe coding' — AI-assisted development workflows where developers describe intent and let AI models generate, refine, and iterate on code within the IDE. The integration

Weekly radar

Product launchMar 25

Major AI Labs Acquire Developer Tooling Companies: OpenAI Buys Astral, Anthropic Buys Bun, DeepMind Acquires Antigravity Team

A clear pattern has emerged among the leading AI labs: each is acquiring established developer tooling companies to strengthen their AI-assisted development ecosystems. OpenAI has acquired Astral (the company behind the popular Python toolchain including the uv package manager and ruff linter),

Weekly radar

Product launchMar 25

Accenture 2026 Research Profile: Consulting Giant's AI and Digital Transformation Strategy

FinancialContent published an in-depth research profile of Accenture (ACN) in 2026, positioning the company as a 'reinvention giant' in the consulting and technology services space. While the full article content was not accessible from the Google News redirect, the framing suggests analysis of Acce

Weekly radar

Business moveMar 25

OpenAI Argues SWE-bench Verified No Longer Measures Frontier Coding Capabilities

OpenAI has published a position stating that SWE-bench Verified, one of the most widely referenced benchmarks for evaluating AI coding agents on real-world software engineering tasks, no longer meaningfully differentiates frontier model capabilities. SWE-bench Verified is a curated subset of the bro

Weekly radar

Business moveMar 25

Opus 4.6, Codex 5.3, and the Post-Benchmark Era in AI Model Evaluation

Interconnects AI examines the latest generation of frontier models — Anthropic's Opus 4.6 and OpenAI's Codex 5.3 — in the context of what the publication describes as the 'post-benchmark era,' where traditional evaluation metrics are increasingly insufficient to capture meaningful differences betwee

Weekly radar

Business moveMar 25

METR Study Examines Task-Completion Time Horizons of Frontier AI Models

METR (Model Evaluation & Threat Research) has published research examining the task-completion time horizons of frontier AI models, measuring how long and how complex the tasks are that current leading AI systems can autonomously complete. This line of research tracks the progression from models tha

Weekly radar

Business moveMar 25

Vibe-Coding Startup Lovable Pursuing Acquisition Strategy

Lovable, a startup in the "vibe-coding" space that enables users to build software applications through natural language prompts rather than traditional coding, is actively seeking acquisition targets. The company operates in the rapidly growing AI-assisted development tools market, where platforms

Weekly radar

Business moveMar 25

Dreamer Acquihired by Meta Superintelligence Labs After 9 Months Building Personal Superintelligence

Latent Space reports that Dreamer, a startup focused on building 'Personal Superintelligence,' has been acquihired by Meta's Superintelligence Labs (MSL), led by Nat and Alex. The deal reportedly happened within days of Dreamer's appearance on the Latent Space podcast, approximately 11 days after re

Weekly radar

Business moveMar 25

Accenture to Acquire Verum Partners to Expand Capital Projects Capabilities in Latin America

Accenture announced its intent to acquire Verum Partners, a consultancy focused on capital projects in Latin America. The acquisition is aimed at expanding Accenture's capabilities in capital project management and delivery across the region. The article content was not fully accessible due to the

Weekly radar

Open sourceMar 25

Hugging Face Surveys 16 Open-Source Reinforcement Learning Libraries for LLM Training

Hugging Face published a comprehensive analysis of 16 open-source reinforcement learning (RL) libraries used for training large language models, focusing on how each approach handles asynchronous token generation and GPU utilization. The blog post, authored by Amine Dirhoussi, Quentin Gallouédec, Ka

Weekly radar

Open sourceMar 25

Opinion Piece Argues Open Source Projects Should Charge for Access

Steven J. Vaughan-Nichols published an opinion piece in The Register arguing that open source maintainers should move beyond donation-based sustainability models and begin charging for access to their code. The article's central thesis is that while a handful of open source developers thrive, most s

Weekly radar

Policy / ResearchMar 25

Deloitte Releases 2026 State of AI in the Enterprise Report

Deloitte has published its 2026 edition of 'The State of AI in the Enterprise,' a recurring benchmark report that tracks how organizations are adopting, scaling, and governing AI across industries. While the full article content was not accessible due to a Google News redirect, Deloitte's annual AI

Weekly radar

Policy / ResearchMar 25

New California AI Laws Take Effect with Compliance Requirements for Businesses

Pillsbury Winthrop Shaw Pittman, a major law firm, has published guidance on new California AI laws that are now in effect or approaching compliance deadlines. While the full article content was not accessible, the title and context indicate that California has enacted a new wave of AI-related legis

Weekly radar

Model releaseApr 7

Google Launches Gemini 3 Pro with Strong Benchmark Results Against GPT-5.1 and Claude Sonnet 4.5

Google has released Gemini 3 Pro, claiming it as the top-performing AI model across multiple benchmarks. According to Google's published results, Gemini 3 Pro surpasses both its predecessor Gemini 2.5 Pro and competitors including OpenAI's GPT-5.1 and Anthropic's Claude Sonnet 4.5. On the ScreenSpot

Weekly radar

Business moveApr 7

Dash0 Reaches Unicorn Status with $110M Series B to Fund Autonomous AI Observability Agents

Dash0, a New York and Berlin-based observability platform founded in 2023, has raised $110 million in a Series B round led by Balderton Capital, with participation from Accel, Cherry Ventures, and Deutsche Telekom's T.Capital. The round values the company at unicorn status and will primarily fund Ag

Weekly radar

Business moveApr 7

U.S. AI Companies Collaborate to Counter Chinese Model Distillation Practices

OpenAI, Anthropic, and Google are collaborating through the Frontier Model Forum to combat adversarial distillation—a practice where Chinese competitors extract outputs from leading U.S. AI models to replicate their capabilities at lower cost. OpenAI has specifically accused Chinese firm DeepSeek of

Weekly radar

Open sourceApr 7

Arcee Releases Trinity Large Thinking, a 400B-Parameter Open-Weight Reasoning Model Built on $20M Budget

Arcee, a 26-person U.S. startup, has released Trinity Large Thinking, a 400-billion-parameter open-weight reasoning model built on a reported $20 million budget. CEO Mark McQuade claims it is the most capable open-weight model ever released by a non-Chinese company, positioning it as a Western alter

Weekly radar

Policy / ResearchApr 7

European Parliament Votes to Delay EU AI Act Implementation

The European Parliament has voted to delay the implementation timeline of the EU AI Act, the landmark comprehensive AI regulation framework. While the full article content was not accessible, the headline indicates a significant shift in the regulatory landscape for AI in Europe, where the AI Act wa

Weekly radar

Model releaseApr 8

Claude Opus 4.6 Surpasses GPT-5.2 on GDPval-AA Agentic Knowledge Work Benchmark

Anthropic's Claude Opus 4.6 has reportedly taken the top position on the GDPval-AA benchmark, which evaluates AI models on agentic real-world knowledge work tasks, surpassing OpenAI's GPT-5.2. The benchmark appears to focus specifically on the ability of models to perform complex, multi-step workflo

Weekly radar

Model releaseApr 8

Anthropic Withholds Claude Mythos Preview from General Release Due to Safety Concerns

Anthropic has released a system card for its latest frontier model, Claude Mythos Preview, confirming earlier leaks about the model's existence. The company describes it as 'by far the most powerful AI model we've ever developed' but has decided not to make it generally available, citing a 'large in

Weekly radar

Product launchApr 8

Harvey AI Announces Autonomous Agents for Legal Industry

Harvey AI, a legal-focused AI company, published a blog post titled "Legal Is Next" announcing its vision and product direction around autonomous AI agents for the legal profession. While the full article content was not available due to the page being rendered client-side, the title and URL strongl

Weekly radar

Product launchApr 8

A2CN: Open Protocol for Agent-to-Agent Commercial Negotiation Launched

A new open protocol called A2CN (Agent-to-Agent Commercial Negotiation) has been introduced, designed to enable autonomous AI agents to conduct B2B commercial negotiations across organizational boundaries. The protocol supports offer exchange, mandate verification, and dual-signed transaction record

Weekly radar

Business moveApr 8

Nvidia-Backed Firmus Reaches $5.5 Billion Valuation in AI Datacenter Expansion

Firmus, an AI datacenter company backed by Nvidia, has reportedly reached a $5.5 billion valuation as it aggressively expands its datacenter footprint to meet surging demand for AI compute infrastructure. The investment from Nvidia signals the chipmaker's continued strategy of investing in downstrea

Weekly radar

Business moveApr 8

Hosted.ai Raises $19M to Address GPU Underutilization in AI Infrastructure

Hosted.ai has raised $19 million in funding to tackle GPU underutilization, a persistent challenge in AI infrastructure economics. The startup is focused on improving the efficiency of GPU resource allocation, addressing the well-documented problem that many organizations deploying AI workloads see

Weekly radar

Model releaseApr 9

Meta Releases Muse Spark, Its First Proprietary AI Model After Shift Away from Open Source

Meta has released Muse Spark, its first major new AI model in over a year, marking a significant strategic shift from the company's previous open-source approach to a proprietary model strategy. The release comes approximately 10 months after Meta spent billions to recruit Scale AI co-founder Alexan

Weekly radar

Business moveApr 9

Snowflake Announces Intent to Acquire Observe for AI-Powered Observability

Snowflake has announced its intent to acquire Observe, an observability platform company, to expand into the $50+ billion IT operations management software market. The acquisition will integrate Observe's observability capabilities directly into the Snowflake AI Data Cloud, enabling enterprises to investigate operational issues using natural language queries across logs, metrics, and traces.

Weekly radar

Business moveApr 9

ServiceNow Acquires Israeli Startup Traceloop for $60-$80 Million in Third Israeli Deal in Three Months

ServiceNow has acquired Traceloop, an Israeli startup focused on observability for LLM-based applications, in a deal valued between $60 and $80 million. This marks ServiceNow's third acquisition of an Israeli company in under three months, signaling an aggressive push to bolster its AI capabilities

Weekly radar

Business moveApr 9

Meta Commits $21 Billion to CoreWeave in Major AI Infrastructure Deal

Meta has announced a $21 billion commitment to CoreWeave, one of the largest single AI infrastructure deals to date. The agreement signals Meta's aggressive strategy to secure GPU compute capacity at massive scale to support its AI research, training, and inference workloads. CoreWeave, a cloud prov

Weekly radar

Business moveApr 9

Anthropic Expands Partnership with Google and Broadcom for Multiple Gigawatts of Next-Generation Compute

Anthropic has announced a significant expansion of its partnership with Google and Broadcom to secure multiple gigawatts of next-generation compute capacity. While full article details were not accessible from the provided content, the title signals a major infrastructure commitment involving custom

Weekly radar

Model releaseApr 10

Zhipu AI Releases GLM-5.1: 754B Parameter Model Tops SWE-Bench Pro Coding Benchmark

Zhipu AI (Z.ai) has released GLM-5.1, a 754-billion parameter large language model that reportedly achieves the top score on the SWE-Bench Pro benchmark, which measures AI systems' ability to resolve real-world software engineering tasks. The model represents a significant scaling step for the Chinese AI ecosystem, competing directly with Western frontier models on code-generation tasks.

Weekly radar

Model releaseApr 10

Meta Superintelligence Labs Announces Muse Spark, First Frontier Model on New Architecture Stack

Meta Superintelligence Labs (MSL), the research division Meta established to pursue artificial superintelligence, has announced Muse Spark — described as its first frontier model built on a completely new architecture stack. This represents MSL's first major public model release since being spun up as a dedicated division, marking Meta's strategic shift away from open-source-only AI models.

Weekly radar

Model releaseApr 10

Alibaba's Qwen Approaches 1 Billion Downloads Following New Model Launch

Alibaba's open-source large language model family, Qwen, is reportedly nearing 1 billion cumulative downloads following the launch of a new model. This milestone positions Qwen as one of the most widely adopted open-weight model families globally, reflecting strong developer interest in alternatives to Western proprietary models.

Weekly radar

Business moveApr 10

Infosys to Acquire Optimum Healthcare IT for $465 Million and Stratus for $95 Million

Infosys has announced plans to acquire two US-based companies: Optimum Healthcare IT for $465 million and Stratus for $95 million. The combined $560 million in acquisitions signals Infosys's continued investment in expanding its US market presence, particularly in healthcare IT and consulting services.

Weekly radar

Business moveApr 10

Anthropic Reaches $30B ARR, Announces Project GlassWing and Claude Mythos Preview

Anthropic has reportedly reached $30 billion in annual recurring revenue, a significant milestone that positions the company as a formidable competitor to OpenAI, particularly as OpenAI navigates its upcoming IPO. The newsletter reports on two major announcements: Project GlassWing, a new research initiative, and Claude Mythos Preview, which Anthropic withheld from general release citing safety concerns.

Weekly radar

Model releaseApr 16

Qwen 3.6 Small MoE Model Running Locally Outperforms Claude Opus 4.7 on SVG Generation Benchmark

Simon Willison, a prominent developer and AI commentator, reported that Alibaba's Qwen3.6-35B-A3B mixture-of-experts model — running locally on his laptop — produced a superior SVG rendering of a pelican riding a bicycle compared to Anthropic's Claude Opus 4.7. The Qwen model uses only 3 billion act

Weekly radar

Product launchApr 16

Z.AI Launches GLM-5.1: Open-Weight 754B Agentic Model with State-of-the-Art SWE-Bench Pro Performance

Z.AI has released GLM-5.1, a 754-billion parameter open-weight model designed specifically for agentic use cases. The model reportedly achieves state-of-the-art results on SWE-Bench Pro, a benchmark focused on real-world software engineering tasks, and is capable of sustaining autonomous execution f

Weekly radar

Product launchApr 16

Shoemaker Allbirds Pivots to AI Data Centers, Stock Surges 580% in One Day

Allbirds, the struggling sustainable footwear and apparel brand, announced a dramatic business pivot away from its core consumer products business toward AI infrastructure. The company plans to sell its existing shoe and apparel operations and use $50 million in financing to reposition itself as a G

Weekly radar

Product launchApr 16

Accenture Acquires Keepler to Strengthen AI and Data Capabilities in Spain

Accenture has announced the acquisition of Keepler, a Spain-based data engineering and AI consultancy, to bolster its AI and data business in the Spanish market. The deal reflects Accenture's ongoing aggressive M&A strategy to expand AI delivery capabilities across European markets, adding specializ

Weekly radar

Business moveApr 16

Cisco to Acquire Galileo to Strengthen AI Observability and Trust in Enterprise Deployments

Cisco has announced its intent to acquire Galileo, a company specializing in AI observability and trust tooling for enterprise deployments. The acquisition signals Cisco's strategic push to embed AI governance, monitoring, and reliability capabilities into its broader enterprise networking and infra

Weekly radar

Open sourceApr 16

Linux Kernel Adds Official Documentation on AI Coding Assistant Usage

The Linux kernel project has added an official documentation file, 'coding-assistants.rst', to its process documentation in the main repository maintained by Linus Torvalds. This file establishes guidelines and expectations for contributors who use AI coding assistants when submitting patches and co

Weekly radar

Policy / ResearchApr 16

AI Vibe Coding Leads to Exposed Patient Data and Privacy Violations in Swiss Medical App

A blog post by Swiss technologist Tobias Brunner details a case where a medical professional used AI-assisted coding tools to build a patient management application without adequate software engineering or security expertise. The resulting application reportedly exposed all patient data unprotected

Weekly radar

Model releaseApr 17

Qwen Team Open-Sources Qwen3.6-35B-A3B Sparse MoE Vision-Language Model with Agentic Coding Capabilities

The Qwen team has open-sourced Qwen3.6-35B-A3B, a sparse Mixture-of-Experts (MoE) vision-language model featuring 35 billion total parameters but only 3 billion active parameters during inference. The model incorporates agentic coding capabilities, positioning it as a compact yet powerful option for

Weekly radar

Model releaseApr 17

Anthropic Releases Claude Opus 4.7 as New State-of-the-Art Model

Anthropic has released Claude Opus 4.7, which the AINews newsletter describes as surpassing its predecessor Claude Opus 4.6 across every evaluated dimension, establishing itself as the new state-of-the-art model. The release continues Anthropic's rapid iteration cadence on its flagship Opus line, wi

Weekly radar

Product launchApr 17

CoreWeave Surges on Meta and Anthropic Deals, Reinforcing Neocloud Dominance in AI Inference

CoreWeave, a GPU-focused cloud infrastructure provider often described as a 'neocloud,' has seen its stock surge following the announcement of major deals with Meta and Anthropic. These contracts position CoreWeave as a critical infrastructure partner for two of the most prominent AI model developer

Weekly radar

Business moveApr 17

Cursor in Talks to Raise $2B at $50B Valuation as Enterprise Growth Surges

Cursor's developer Anysphere is reported to be in talks to raise $2 billion at a $50 billion valuation, driven by surging enterprise adoption of its AI-powered code editor. The round, if completed, would make Cursor one of the most valuable private AI companies globally. The valuation reflects intense enterprise demand for AI coding tools and signals that the market views Cursor as a durable category leader, not just a consumer novelty.

Weekly radarcursor

Business moveApr 17

Factory Raises $150M at $1.5B Valuation to Scale AI Coding Agents for Enterprises

Factory raised $150 million at a $1.5 billion valuation to scale its AI coding agent platform for enterprise customers. The company focuses on autonomous software development workflows, targeting engineering teams looking to increase throughput beyond what copilot-style assistants offer. The round reinforces investor conviction in purpose-built agentic coding platforms as a distinct and durable category separate from general-purpose AI assistants.

Weekly radarfactory

Pricing / PlatformApr 17

Meta Signs $21 Billion AI Infrastructure Contract with CoreWeave

Meta has signed a $21 billion AI infrastructure contract with CoreWeave, marking one of the largest single deals in the rapidly expanding AI compute market. The deal underscores the massive capital expenditures hyperscalers and major tech companies are committing to secure GPU compute capacity for t

Weekly radar

Model releaseApr 18

Anthropic Invests in Hardware Design Startup Building 'Cursor for Hardware'

WIRED reports that Anthropic is investing in a startup that has built what is being described as the 'Cursor for Hardware' — an AI-assisted design and development tool aimed at hardware engineering workflows. The investment signals Anthropic's interest in expanding Claude's applicability beyond soft

Weekly radar

Business moveApr 18

Cursor Reportedly Raising Over $2 Billion at $50 Billion Valuation Amid Enterprise Growth

Cursor, the AI-powered code editor built by Anysphere, is reportedly in talks to raise more than $2 billion in new funding at a valuation of $50 billion, according to TechCrunch sources. The round would represent a dramatic increase in the company's valuation, driven by surging enterprise adoption o

Weekly radar

Model releaseApr 19

Figma's Business Model Faces Growing Pressure from Claude Design Launch

Martin Alderson argues that Figma's business model is uniquely vulnerable to AI disruption, specifically citing the launch of Anthropic's Claude Design as a compounding threat. The core thesis is that Figma's revenue has historically depended heavily on non-designer seats — product managers, enginee

Weekly radar

Product launchApr 19

xAI Plans to Supply Computing Power to AI Coding Startup Cursor

Elon Musk's xAI is reportedly planning to supply computing power to Cursor, the AI-assisted coding startup developed by Anysphere. This deal would see xAI leveraging its substantial GPU infrastructure — including its Memphis-based Colossus supercomputer cluster — to serve as a compute provider for o

Weekly radar

Model releaseApr 21

Moonshot Releases Kimi K2.6 Open Model Approaching Anthropic Opus 4.6 Performance

Moonshot AI has released Kimi K2.6, an update to its open-weight model line that reportedly approaches the performance of Anthropic's Claude Opus 4.6 on key benchmarks. The release positions Kimi K2.6 as one of the leading open models available, with speculation that it may be timed ahead of an anti

Weekly radar

Model releaseApr 21

Meta Plans to Open Source Versions of Its Next AI Models

According to an Axios scoop, Meta is preparing to open source versions of its next generation of AI models. While the full article content was not accessible beyond the headline, the report signals Meta's continued commitment to its open-weight model strategy, which has been a defining element of th

Weekly radar

Model releaseApr 22

OpenAI Launches GPT-Image-2 Image Generation Model

OpenAI has launched GPT-Image-2, its next-generation image generation model. The release represents a continued investment by OpenAI in multimodal capabilities, building on the image generation features integrated into GPT-4o and subsequent models. Details on specific architectural improvements and

Weekly radar

Model releaseApr 22

Mozilla Uses Claude to Identify 271 Security Vulnerabilities in Firefox 150

Mozilla's ongoing collaboration with Anthropic has yielded a significant security milestone: an early version of Claude Mythos Preview was applied to the Firefox codebase, resulting in the identification of 271 vulnerabilities that are patched in this week's release of Firefox 150. The findings were

Weekly radar

Model releaseApr 22

Qwen 3.6-27B Delivers Flagship-Level Agentic Coding in a Dense 27B Model

Alibaba's Qwen team has released Qwen3.6-27B, a 27-billion parameter dense model that they claim surpasses their previous-generation flagship Qwen3.5-397B-A17B (a 397B total / 17B active mixture-of-experts model) across agentic coding benchmarks. The model represents a significant efficiency breakth

Weekly radar

Business moveApr 22

SpaceX Secures Option to Acquire AI Coding Startup Cursor for $60 Billion

According to The Wall Street Journal, SpaceX has secured an option to acquire Anysphere, the company behind the AI-powered code editor Cursor, for approximately $60 billion. Cursor has rapidly emerged as one of the most prominent AI-assisted development tools, competing directly with GitHub Copilot,

Weekly radar

Model releaseApr 23

Shopify CTO Details AI Phase Transition: Unlimited Opus-4.6 Token Budget, Internal Tools Tangle, Tangent, and SimGym

In a detailed interview on the Latent Space podcast, Shopify CTO Mikhail Parakhin revealed the scope of Shopify's AI transformation in 2026, describing it as a 'phase transition' in internal AI usage. Parakhin disclosed that Shopify has adopted an unlimited token budget for Anthropic's Opus-4.6 mode

Weekly radar

Model releaseApr 23

Alibaba Qwen Team Releases Qwen3.6-27B Dense Model Outperforming 397B MoE on Agentic Coding Benchmarks

Alibaba's Qwen team has released Qwen3.6-27B, a dense open-weight language model with 27 billion parameters that reportedly outperforms much larger mixture-of-experts (MoE) models — specifically a 397B parameter MoE — on agentic coding benchmarks. The model represents a significant efficiency breakt

Weekly radar

Business moveApr 23

Nvidia Backs Vast Data at $30 Billion Valuation in AI Infrastructure Bet

Nvidia has invested in Vast Data, a data infrastructure company, at a reported $30 billion valuation. The investment signals Nvidia's continued strategy of backing companies across the AI infrastructure stack beyond its core GPU business, extending into data platforms that support AI workloads. Vast

Weekly radar

Business moveApr 23

Microsoft Explored Acquiring Cursor Before SpaceX Deal

According to CNBC sources, Microsoft explored acquiring Cursor, the AI-powered code editor developed by Anysphere, before the startup ultimately pursued a deal involving SpaceX. The report highlights Microsoft's strategic interest in consolidating the AI-assisted development tooling market, where it

Weekly radar

Business moveApr 23

Meta Commits $21B to CoreWeave Through 2032, Signaling Neoclouds as Core AI Infrastructure

Meta has signed a $21 billion agreement with CoreWeave, the GPU-specialized neocloud, through December 2032. The deal provides Meta with distributed AI cloud infrastructure featuring NVIDIA Vera Rubin GPUs across multiple locations. CoreWeave CEO Michael Intrator stated: "This is another example tha

Weekly radar

Open sourceApr 23

Tencent Cloud Open-Sources CubeSandbox: Sub-60ms Micro-VM Sandbox Using RustVMM and KVM

Tencent Cloud has open-sourced CubeSandbox, a lightweight micro-VM sandbox built on RustVMM and KVM that claims cold-start times under 60 milliseconds. The project positions itself as an open-source alternative to E2B, the commercial sandboxing platform widely used for running AI-generated code in i

Weekly radar

Model releaseApr 24

GPT-5.5 and OpenAI Codex Superapp Launch: Computer-Use, Document Automation, and Overnight Runs

OpenAI launched GPT-5.5 on April 24, 2026, a model that emphasizes cost-efficiency rather than raw capability alone, delivering "better long-horizon execution, stronger computer-use behavior, and materially improved token efficiency." API pricing is $5/$30 per million input/output tokens with a Pro

Weekly radar

Model releaseApr 24

DeepSeek V4: Frontier-Class Open-Weight Model with Million-Token Context at a Fraction of the Price

DeepSeek released two Mixture-of-Experts model checkpoints on April 24, 2026: V4-Pro (1.6T total parameters, 49B active) and V4-Flash (284B total, 13B active), both with 1 million token context windows under an MIT license. The pricing is dramatically competitive — V4-Flash costs $0.14 per million i

Weekly radar

Business moveApr 24

Google Plans $40 Billion Investment in Anthropic at $350B Valuation

Google announced it will invest up to $40 billion in Anthropic at a $350 billion valuation, with $10 billion committed immediately in cash. The move solidifies Google as Anthropic's largest external investor and comes alongside a separate April 6 deal where Anthropic signed its largest-ever compute

Weekly radar

Business moveApr 25

Cohere Acquires Aleph Alpha in $20B Transatlantic Sovereign-AI Deal Anchored by Schwarz Group's $600M

Canada's Cohere announced on April 24 it will merge with Germany's Aleph Alpha at a combined valuation of approximately $20 billion, anchored by a $600 million Series E commitment from the Schwarz Group (Lidl, Kaufland, Schwarz Digits). The deal — described by Cohere CEO Aidan Gomez as complementary

Weekly radar

Business moveApr 25

Cognition AI in Talks at $25B Valuation as Devin/Windsurf ARR Doubles to ~$1B

Cognition AI is in funding talks at a $25 billion valuation, more than 2x its September 2025 $10.2B mark, on the back of an estimated ARR jump from ~$400M to $800M-$1B since acquiring Windsurf in July 2025. Bloomberg reported the talks on April 25, 2026; the round is unconfirmed but possible leads i

Weekly radardevin-windsurf-cursor-github-copilot

Product launchApr 28

NVIDIA Releases Nemotron 3 Nano Omni: 30B Multimodal MoE Supporting Documents, Audio, and Video

NVIDIA released Nemotron 3 Nano Omni on April 28, 2026, a 30-billion-parameter multimodal mixture-of-experts model supporting text, image, video, audio, and document inputs with a 256K context window. The model saw immediate adoption across Ollama, LM Studio, and OpenRouter within hours of release,

Weekly radar

Business moveApr 28

OpenAI Misses Q1 2026 Revenue Targets as Anthropic Overtakes in Run-Rate Revenue

OpenAI fell short of multiple internal revenue and user growth targets in Q1 2026, triggering a sharp market selloff in AI infrastructure stocks including Oracle (-7.7%), CoreWeave (-7.4%), SoftBank (-10%), and Nvidia (-2-6%). The company missed its internal goal of one billion weekly active ChatGPT

Weekly radar

Open sourceApr 29

Warp Terminal Client Open-Sourced Under AGPLv3 with OpenAI as Founding Sponsor

Warp released its terminal client as open source under the AGPLv3 license, with OpenAI joining as a founding sponsor alongside contributions from the Oz running agent project. The move signals a strategic shift toward community-driven development while retaining a commercial offering. Open-sourcing a polished AI-native terminal increases competitive pressure on proprietary alternatives and could accelerate adoption of agentic development workflows.

Weekly radarwarp

ShutdownApr 30

Amazon Q Developer End-of-Support Announced — Signups Block May 2026, Full EOS April 2027

AWS announced end-of-support for Amazon Q Developer, with new signups blocked starting May 15, 2026, and full end-of-support on April 30, 2027. The announcement marks a significant retreat from the competitive AI coding assistant market, where GitHub Copilot, Cursor, and others have established dominant positions. Existing Q Developer customers will need to migrate, creating a near-term opportunity for competing tools and partners building on those platforms.

Weekly radaramazon-q-developer-kiro

Business moveMay 4

Sierra Raises $950M at $15.8B Valuation as Enterprise AI Agent Demand Surges

Sierra, the enterprise conversational AI company founded by Bret Taylor (OpenAI chairman, former Salesforce co-CEO) and Clay Bavor (former Google), raised $950 million in a round led by Tiger Global and Google's GV, pushing its valuation to $15.8 billion. The funding follows rapid ARR growth from $1

Weekly radar

Business moveMay 4

Anthropic and OpenAI Launch Rival $1.5B and $10B Enterprise AI Joint Ventures with Wall Street

Anthropic and OpenAI simultaneously announced competing joint ventures to accelerate enterprise AI deployment through Wall Street-backed financing structures, dropping on May 4 within hours of each other. Anthropic's venture — unnamed, valued at $1.5 billion — is backed by Blackstone, Hellman & Frie

Weekly radarclaude-code-github-copilot

Model releaseMay 5

Latent Space: GPT-5.x Derives Novel Theoretical Physics — Vibe Physics and the Science Frontier

Latent Space's May 5 podcast episode featuring physicist Alex Lupsasca documents a qualitative shift in AI capability: GPT-5.x models are generating original theoretical physics research, not reproducing existing work. Lupsasca, a prize-winning theoretical physicist at OpenAI, describes two specific

Weekly radar

Business moveMay 5

Anthropic Targets Midmarket Software Spend with Custom Claude Systems as SaaS Replacement

The Register's coverage of Anthropic's new enterprise-services firm (backed by Blackstone, Hellman & Friedman, Goldman Sachs) frames the play explicitly as a midmarket SaaS-replacement strategy. Anthropic engineers will collaborate with the new firm's staff to build custom Claude-powered systems for

Weekly radarclaude-code

Business moveMay 6

Anthropic Secures 300MW SpaceX Colossus Compute Deal, Doubles Claude Code Rate Limits

Anthropic has signed a compute partnership with SpaceX to access all capacity at the Colossus 1 data center in Memphis, Tennessee — more than 300 megawatts and approximately 220,000 NVIDIA GPUs — with deployment expected within a month. The deal complements existing agreements with Amazon (up to 5 G

Weekly radar

Business moveMay 6

Latent Space: Silicon Valley Gets Serious About Services — Anthropic and OpenAI Launch Enterprise Deployment Ventures

Latent Space's May 6 AI News edition analyzes the week's defining development: both Anthropic and OpenAI launched enterprise services businesses in the same week, signaling a fundamental shift in how frontier AI labs see the path to revenue at scale. Anthropic's $1.5B joint venture with Blackstone,

Weekly radar

Model releaseMay 8

Anthropic Introduces Natural Language Autoencoders to Read Claude's Internal Thinking

Anthropic has published research on Natural Language Autoencoders (NLAs), a technique that converts the numerical activations inside Claude's hidden layers directly into natural-language text explanations that anyone can read. The system uses a two-component architecture: an Activation Verbalizer (A

Weekly radar

Business moveMay 9

Anthropic Growing 10x/Year, Reaches $1–1.2T Valuation as Tech Sector Restructures Around AI

Latent Space's May 9 AI News edition documents a sharp divergence between Anthropic's explosive growth and widespread restructuring across the broader tech sector. Anthropic is experiencing 10x annual growth, has reached a $1–1.2 trillion valuation (making it the 11th–15th most valuable company glob

Weekly radar

Product launchMay 11

Microsoft Copilot Studio April Release Ships Agent Governance, MCP Connectors, and App Surfaces

Microsoft's April 2026 Copilot Studio release (posted May 11) ships three coordinated capability sets: agent governance, intelligent workflows, and connected app experiences. The Analytics Viewer admin role (now GA) grants read-only access to agent analytics for cost and operational oversight withou

Weekly radargithub-copilot

Business moveMay 11

OpenAI Deployment Company + Tomoro Acquisition Hit Accenture, Cognizant, Infosys Shares

On May 11, 2026, OpenAI announced the formation of an OpenAI Deployment Company and the acquisition of AI engineering firm Tomoro, with the stated goal of helping enterprises take OpenAI capabilities to production. Markets reacted as if this were a direct shot at the IT services majors: Accenture (A

Weekly radarcodex

Product launchMay 12

Glean Introduces Enterprise Agent Development Lifecycle to Govern AI Agent Sprawl

Glean unveiled its Enterprise Agent Development Lifecycle (ADLC), a seven-stage framework (Opportunity, Design, Performance, Input, Develop, Launch, Monitor & Improve) designed to give enterprises a repeatable methodology for shipping AI agents in production. The framework progresses from identifyin

Weekly radar

Business moveMay 14

Anthropic Overtakes OpenAI Among Business Customers as It Courts Small Firms and Eyes $950B Valuation

Anthropic has surpassed OpenAI in verified business customer share for the first time, with 34.4% of businesses paying for Anthropic versus 32.3% for OpenAI — a dramatic 12-month surge from just 9% adoption in May 2025. The shift is concentrated in technical sectors (finance, technology, professiona

Weekly radarclaude-code

Model releaseMay 18

Alibaba releases Qwen3 Coder Next — open-source code specialist tuned for software engineering

Alibaba's Qwen team released Qwen3 Coder Next on May 18, a next-generation code-specialist variant tuned for software engineering benchmarks and agentic coding loops. Availability surfaced on LLM Gateway and the Qwen3 family hub. The release follows Qwen 3.5 (February, 397B MoE with unified vision-l

Weekly radar

Model releaseMay 18

Cursor releases Composer 2.5 — matches Claude Opus 4.7 on coding at one-tenth the cost

Cursor shipped Composer 2.5 on May 18, its in-house long-horizon coding model built on Moonshot's open-source Kimi K2.5 checkpoint with 25× the synthetic training task volume of Composer 2. The model scores 79.8% on SWE-Bench Multilingual and 63.2% on CursorBench v3.1 — matching Claude Opus 4.7 and

Weekly radar

Product launchMay 18

IBM launches Open Agent Leaderboard for benchmarking enterprise agentic systems

IBM Research published the Open Agent Leaderboard on Hugging Face on May 18 — a benchmark suite designed to evaluate enterprise agentic systems across multi-step task completion, tool use accuracy, governance compliance, and cost-per-task. The leaderboard formalizes evaluation across major open and

Weekly radar

Business moveMay 18

Anthropic acquires Stainless, the SDK and MCP tooling startup used by OpenAI and Google

Anthropic announced on May 18 that it has acquired Stainless, the New York–based developer tools startup behind the SDK generators used by OpenAI, Google, Cloudflare, and most major MCP server implementations. Terms were not disclosed; The Information reported the deal was in the $300M+ range. Stain

Weekly radar

Model releaseMay 19

OpenAI partners with Dell to bring Codex into hybrid and on-premises enterprise environments

OpenAI and Dell Technologies announced on May 19 a partnership to bring Codex into hybrid and on-premises enterprise environments. The deal follows the OpenAI Deployment Company launch (Tomoro acquisition, $4B+ in PE backing from TPG, Brookfield, Advent, Bain) and is positioned as the on-prem comple

Weekly radar

Model releaseMay 19

KPMG integrates Claude across 276,000-person workforce in Anthropic strategic alliance

Anthropic and KPMG announced a strategic alliance on May 19 that integrates Claude across KPMG's core business and global workforce of more than 276,000 people. KPMG joins PwC (which announced earlier in May it is deploying Claude to build technology, execute deals, and reinvent enterprise functions

Weekly radar

Model releaseMay 19

Andrej Karpathy joins Anthropic to lead Claude pre-training research team

Andrej Karpathy — OpenAI co-founder, former Tesla AI lead, and one of the most-followed practitioner voices in deep learning — has joined Anthropic, where he will start a team focused on using Claude to accelerate pre-training research. He reports to pre-training lead Nick Joseph. Karpathy framed th

Weekly radar

Policy / ResearchMay 19

OpenAI Joins C2PA, Adds SynthID Watermarks, Launches Public Image Verification Tool

OpenAI announced on May 19, 2026 that it has become a C2PA Conforming Generator Product, integrated Google DeepMind's SynthID invisible watermarking into images generated by ChatGPT, Codex, and the OpenAI API, and launched a preview public verification tool that lets anyone check whether an image wa

Weekly radar

Model releaseMay 20

Mistral AI secures €200M to scale open-weights models

France's Mistral AI secured €200 million in new funding announced May 20, with proceeds earmarked to scale Mistral's open-weights model family. The round follows Mistral Medium 3.5's late-April release and Mistral Large 3 / Small 4 moving under Apache 2.0 licensing. With Llama 4 (Meta), Qwen 3.5/3.6

Weekly radar

Product launchMay 20

Google Launches Antigravity 2.0 — Multi-Agent Suite with Subagents, Browser, CLI, and Custom-Agent SDK

Google launched Antigravity 2.0 at Google I/O on May 19, 2026, positioned as a multi-agent development suite rather than a code editor. The defining capabilities are coordinated multi-agent workflows — dynamic subagents that work in parallel across projects, scheduled background tasks for autonomous

Weekly radarantigravity-cursor-windsurf-claude-code-github-copilot

Product launchMay 20

Harvey partners with DeepJudge to ground legal AI workflows in firms' institutional knowledge

Harvey, the dominant generative-AI platform for law firms, announced on May 20 a partnership with DeepJudge — the ex-Google interpretability-veteran-founded legal AI startup — to bring DeepJudge's institutional-knowledge graph into Harvey's workflows. DeepJudge surfaces firms' past work, drafting de

Weekly radar

Business moveMay 21

Anthropic in Early Talks to Run Claude on Microsoft's Custom Maia 200 AI Chips

Anthropic is in early-stage talks with Microsoft to rent Azure servers powered by Microsoft's custom Maia 200 AI accelerator, according to reporting first surfaced by The Information on May 21 and substantiated by CNBC and Reuters. The discussions follow Microsoft's $5 billion investment in Anthropi

Weekly radarclaude-code

Business moveMay 21

Anthropic Makes First Acquisition — Fractional AI — to Anchor Enterprise Services Venture

Anthropic completed its first acquisition, naming San Francisco-based Fractional AI as the operational foundation of its newly launched enterprise consulting venture. The venture — backed by Blackstone, Hellman & Friedman, Apollo Global Management, General Atlantic, GIC, and Sequoia Capital — has co

Weekly radarclaude-code

Business moveMay 22

New AI Infrastructure Unicorns: Exa, Modal, and TurboPuffer Cross Billion-Dollar Threshold

Latent Space's AINews highlighted three AI infrastructure companies that have crossed unicorn status in May 2026, all targeting different layers of the AI application stack. TurboPuffer reached $100M ARR with under $1M raised total and is profitable — positioning the company as the magnet for contex

Weekly radar

Business moveMay 27

Latent.Space: Inference Layer Mints Decacorns — Fireworks at $15B, Baseten at $11B, OpenRouter Raises $113M

Five days after Latent.Space mapped a wave of AI infrastructure unicorns (Exa, Modal, TurboPuffer), three more inference-layer vendors crossed or approached the $10B "decacorn" line in the same week. Fireworks is in talks for a round that would value the company at ~$15B — a 3.75x markup in seven mo

Weekly radar

Business moveMay 28

Cognition Raises $1B Series D at $26B Valuation — Now Largest Independent Agent Lab at $492M ARR

Cognition AI closed a $1 billion Series D at a $26 billion post-money valuation on May 28, co-led by Lux Capital, General Catalyst, and 8VC. The valuation marks a 2.5x step-up from the $10B Series C in September 2025. The company now runs at a $492 million annualized revenue run-rate, has grown ente

Weekly radardevin-windsurf

Business moveMay 28

Anthropic Closes $65B Series H at $965B Valuation, Eclipsing OpenAI as Most Valuable AI Company

Anthropic has closed a $65 billion Series H round at a $965 billion post-money valuation, led by Altimeter, Dragoneer, Greenoaks, and Sequoia. The round vaults Anthropic past OpenAI (valued at roughly $852 billion in late March 2026) to become the most valuable AI company, and brings it to the thres

Weekly radarclaude-code

Business moveMay 29

Anthropic's run-rate revenue hits $47 billion, up from $14B in February

Anthropic disclosed in its Series H funding announcement that run-rate revenue reached $47 billion as of early May 2026 — defined as the most recent month's revenue annualized (×12). The trajectory is steep: $14B in February 2026, ~$30B by early April, and $47B in May. The figures come from Anthropi

Weekly radarclaude-code

Model releaseJun 1

JetBrains Releases Mellum2, a 12B MoE Open Model for High-Frequency Coding Sub-Tasks

JetBrains released Mellum2, an open-source (Apache 2.0) mixture-of-experts model with 12 billion total parameters and 2.5 billion active per token, optimized for text and code workloads. The MoE design keeps total capacity high while activating a small subset per token, delivering what JetBrains rep

Weekly radar

Business moveJun 1

Anthropic Confidentially Files for IPO at ~$1 Trillion Valuation

Anthropic confidentially filed for an initial public offering with the SEC on June 1, 2026, the most concrete signal yet that the Claude maker intends to test public markets while the AI build-out is still in full swing. The confidential route lets Anthropic gather regulatory feedback before any pub

Weekly radar

Model releaseJun 2

Microsoft Launches Seven In-House MAI Models at Build 2026, Led by MAI-Thinking-1, to Cut OpenAI Reliance

At Build 2026 (June 2), Microsoft AI unveiled seven self-developed "MAI" models spanning text, image, voice, and speech, headlined by MAI-Thinking-1 — its first in-house reasoning model. MAI-Thinking-1 is a sparse Mixture-of-Experts model with ~35B active / ~1T total parameters and a 256K-token cont

Weekly radar

Policy / ResearchJun 9

Cognition Launches FrontierCode, Benchmarking Whether AI Code Would Actually Be Merged

Cognition (maker of Devin) introduced FrontierCode, a coding benchmark — explicitly modeled on FrontierMath — that measures whether AI-generated code would actually be merged into production, not merely whether it passes unit tests. Each task required 40+ hours of work by open-source maintainers a

Weekly radardevin

Model releaseJun 10

Anthropic Launches Claude Fable 5, a Publicly Available Mythos-Class Model

Anthropic released Claude Fable 5, its newest publicly available "Mythos"-class model, with reported benchmark wins across coding and reasoning. Early hands-on impressions (Simon Willison, "Initial impressions of Claude Fable 5," 2026-06-09) describe it as a powerful model with broad knowledge and s

Weekly radarclaude-code

Model releaseJun 11

Simon Willison: Claude Fable 5 Is "Relentlessly Proactive" in Agentic Debugging

Simon Willison documents a striking behavioral trait of Anthropic's newly public Claude Fable 5: given a minimal prompt about a stubborn Datasette CSS scrollbar bug, the model independently invented an elaborate multi-tool debugging campaign without being told to. It opened Firefox and Safari window

Weekly radarclaude-code

Model releaseJun 12

US Government Directive Forces Anthropic to Suspend Claude Fable 5 and Mythos 5 for Foreign Nationals

On June 12, 2026, Anthropic received a US government directive ordering it to suspend all access to its two most capable models — Claude Fable 5 and the limited-availability Claude Mythos 5 — "by any foreign national, whether inside or outside the United States, including foreign national Anthropic

Weekly radarclaude-code

Product launchJun 15

OpenAI Launches $150M Partner Network, Escalating the Enterprise Implementation Land Grab

OpenAI launched the OpenAI Partner Network on June 14, 2026, committing $150 million to help consulting firms and systems integrators build and deploy enterprise AI solutions. The program uses a three-tier structure (Select, Advanced, Elite) gated on sales performance, technical capability, and depl

Weekly radar

Business moveJun 15

Accenture's June 18 Earnings Become the Market Test of Whether AI Consulting Is Real

Ahead of Accenture's June 18 earnings, analysts have coalesced around three metrics that will reveal whether AI consulting is genuine structural growth or a crested wave: (1) the bookings trajectory — whether AI-related bookings keep accelerating or have plateaued; (2) AI consulting revenue actually

Weekly radar

Model releaseJun 17

Z.ai's Open-Weights GLM-5.2 Beats GPT-5.5 on Long-Horizon Coding Benchmarks at One-Sixth the Cost

Chinese lab Z.ai (formerly Zhipu AI) released GLM-5.2, a 753B-parameter mixture-of-experts model (≈40B active params, 1.51TB) engineered for long-horizon autonomous coding and engineering tasks. Core weights shipped under an unrestricted MIT license on June 16 following a June 13 subscription launch

Weekly radar

Business moveJun 18

Accenture's Worst-Ever Day: Shares Crash ~20% as AI Disruption Hits Consulting Demand, $4.18B Cybersecurity Pivot

On June 18, 2026, Accenture had the worst single trading day in its history as a public company, with shares falling roughly 20% — leaving the stock down more than 50% for the year. Fiscal Q3 2026 results were technically positive (revenue $18.7B, up 6%; EPS $3.80, up 9%) but new bookings fell ~2% y

Weekly radarcodex

Model releaseJun 26

OpenAI Previews GPT-5.6 (Sol, Terra, Luna) — Staggered Release Gated Behind US Government Pre-Deployment Review

OpenAI began a limited preview of GPT-5.6 on June 26, 2026 — a three-tier model family: Sol (flagship, tuned for ambitious agentic work; $5/1M input, $30/1M output), Terra (balanced everyday model; $2.50/$15), and Luna (fast, high-volume; $1/$6). The release also ships more predictable p

Weekly radar

Pricing / PlatformJun 30

Anthropic Launches Claude Sonnet 5 as a Cheaper Way to Run Agents

Anthropic released Claude Sonnet 5 on June 30, 2026, a midsize model positioned to deliver agentic capabilities — autonomous planning, tool use (browsers, terminals), and extended multi-step task execution — that previously required larger, more expensive frontier models. Anthropic frames the releas

Weekly radar

Product launchJul 2

Microsoft Launches "Frontier Company" — $2.5B Bet on AI Deployment, Not Just AI Tools

Microsoft unveiled Microsoft Frontier Company, a new operating business dedicated to delivering enterprise AI deployments using Microsoft's own AI stack, backed by a $2.5 billion investment and roughly 6,000 industry and engineering experts. Commercial Business CEO Judson Althoff described it as goi

Weekly radar

Model releaseJul 8

OpenAI Introduces GPT-Live — Voice Model That Delegates Hard Tasks to a Frontier Model Mid-Conversation

OpenAI announced GPT-Live on July 8, 2026, an upgraded model powering ChatGPT's voice mode and replacing the previous GPT-4o-era voice system. The distinguishing behavior is task delegation: for questions that require web search, deeper reasoning, or more complex work, GPT-Live hands off to OpenAI's

Weekly radar

Product launchJul 8

NVIDIA and Hugging Face Release Open Synthetic Data for Training AI Agents

NVIDIA released a set of open synthetic datasets and tooling on Hugging Face aimed at training AI agents for real-world complexity rather than benchmark performance — the stated premise being that "the real world does not behave like a benchmark," so agents need data that teaches them to recover fro

Weekly radar

Model releaseJul 9

OpenAI Ships GPT-5.6 (Sol, Terra, Luna) to General Availability — Sol Beats Claude Fable 5 on Agents' Last Exam

On July 9, 2026, OpenAI released the GPT-5.6 family to general availability across ChatGPT, Codex, and the API — moving beyond the June 29 limited preview. The family has three tiers: Sol (flagship, complex reasoning and agentic workloads, $5 in / $30 out per 1M tokens), Terra (balanced, com

Weekly radar