In this edition, covering 3 to 9 November:
- OpenAI will acquire Pierre Computer and add its Code Storage function to Codex, and JPMorgan Chase will use Codex and OpenAI's cybersecurity models.
- Meta acquired Conductor, which helps developers run multiple AI coding agents at the same time, and Mistral AI acquired French AI coding startup SkipLabs.
- Anthropic released Claude Fable 5.2 and Mythos 5.2 with improved coding, reasoning and vision and outputs up to 30% faster, with Fable 5.2 priced at $8 per million input tokens and $40 per million output tokens.

This spot is open for 2027. One company puts its name on every AI Programming Weekly, which has 3,800+ subscribers. See what's included
OpenAI
OpenAI will acquire developer tools startup Pierre Computer and integrate its Code Storage function into Codex; Pierre's open source projects stay available and terms were not disclosed. JPMorgan Chase will use OpenAI's Codex coding agent and cybersecurity models through the Daybreak program, and neither company disclosed financial terms or the agreement's duration. Pierre Computer · JPMorgan Chase
Cat Wu, who led product for Claude Code at Anthropic alongside creator Boris Cherny, left the company and joined OpenAI to lead product management for its core products team. Cat Wu leaves Anthropic · Cat Wu joins OpenAI
DevRev uses OpenAI models for Computer and Operator, tools that handle customer and employee requests with company data and permissions, and uses Codex for its own software development. Rocky Linux creator Kurtzer built Ginux, a Linux compatible kernel written mostly in Rust, in 11 days through a single Codex task; it remains a research project without an independent security audit. DevRev · Ginux
Anthropic
Anthropic released version 5.2 of Claude Fable and Mythos, with improved coding, reasoning and vision and outputs up to 30% faster than the 5.1 models. Fable 5.2 is priced at $8 per million input tokens and $40 per million output tokens, while Mythos 5.2 remains restricted to approved cybersecurity and biomedical research programs. IBM plans to integrate Claude Mythos cyber reasoning into Concert Protect to help enterprise security teams find unknown vulnerabilities, assess exploitability and speed up fixes, with developer approval required for generated fixes. Claude 5.2 models · IBM Concert Protect
Acquisitions and funding
Meta acquired Conductor, a startup founded in 2024 that helps developers run and manage multiple AI coding agents at the same time, for undisclosed terms. Mistral AI acquired SkipLabs, a French startup building tools for faster, more reliable and secure AI code generation, in its fifth acquisition of 2026. Cognizant completed its acquisition of Wizeline, an AI focused digital engineering and product development firm; financial terms were not disclosed. Meta and Conductor · Mistral and SkipLabs · Cognizant and Wizeline
Copado acquired Valo.ai, which helps Salesforce teams find dormant integrations, excessive access and other compliance risks, and adds it to its AgentOps platform. e2e, based in San Francisco and Warsaw, raised €7 million in seed funding led by Credo Ventures, KAYA, Point Nine, Baobab Ventures and Script Capital for an open source AI testing framework. Phaze, founded by the creators of Parsec, raised an $8.75 million seed round led by a16z for remote computers that people and AI agents use to run software. Copado and Valo.ai · e2e · Phaze
Coding agents and app builders
Cursor introduced Build Live Apps on paid plans, where a cloud agent builds a described web app, sets up a database and authentication, runs security checks and deploys it to a shareable link. FlutterFlow introduced an AI native builder that creates and deploys native iOS and Android apps through conversation, while its visual product continues as FlutterFlow Studio. Replit opened its first Asian office in Tokyo on 4 November, and Japanese customer Sumitomo Mitsui Finance and Leasing is expanding its partnership with the company. Automattic launched Spacefast, a service that hosts vibe coded apps. Cursor · FlutterFlow · Replit · Spacefast
Tessl launched a platform where teams build reusable agent skills and combine them into workflows called loops, and says its cost per pull request fell from $75 to $13. Sinch joined the SpaceX Super Intelligence Partner Network, and its Cursor Marketplace plugin gives Grok Build developers Sinch voice, messaging, verification and email tools. Ricursive Intelligence released Ripple-65B Preview, an open coding model by former Google employees, under Apache 2.0 and claims it beats larger US open weight models on two coding benchmarks. Tessl · Sinch · Ripple-65B
ExpertCrowd and HUD released a software engineering evaluation suite that combines tasks and grading rubrics from Toptal experts with HUD's reinforcement learning infrastructure. Amazon revised its internal software engineering guidance to put customer and business outcomes ahead of code volume or AI usage measures, and AWS created Velocity Labs in Seattle for rapid AI prototypes. AWS launched Semarak AI Malaysia, which includes GovHub, powered by Kiro, with a pilot for 500 government developers in 2026 and a goal of 4,000 by 2028. ExpertCrowd SWE · Amazon guidance · Semarak AI Malaysia
DeepSeek Harness, its open source agent framework, reached 1.54 million daily users on DeepSeek's API, up from 429,000 in early October, with about 35,000 third party plugins. DeepSeek Harness
Vulnerabilities in coding agents
Mistral Vibe versions before 3.0.3 have a remote code execution flaw rated 8.6 under CVSS 4.0, in which repository git settings run an attacker controlled program before the trust prompt appears. Docker Sandboxes 0.51.0, released on 3 November, fixes CVE-2026-103916, which let AI agents use symbolic links in the sbx cp command to reach files outside approved folders, including SSH keys. Check Point Research found that Haiou, a Chinese Codex jailbreak bundle hosted on GitHub, writes persistent advertising for an LLM gateway into global agent instruction files. Mistral Vibe · Docker Sandboxes · Haiou
Claude Code's "don't ask again" option saved approved commands, secrets included, to a project settings file that leaked 2,539 developer secrets to public GitHub repositories, Apiiro found; Anthropic says it is looking at fixes. Claude Code secrets
Security tools and research
Google Cloud added CodeMender features for isolated vulnerability testing, exploit simulation, human approval checks and targeted repository scanning, plus agent controls for identity and network access. SecureFlag analyzed more than 800,000 attempts to fix software vulnerabilities and found that 25% were implemented incorrectly, especially for broken access control and data integrity issues. Constructive released Safegres, an MIT licensed tool that audits PostgreSQL grants and row level security policies in read only mode and can run in CI. CodeMender · SecureFlag · Safegres
Checkmarx is now in Anthropic's Claude Marketplace, so US customers can put part of their Anthropic spending commitment toward Checkmarx Fusion SAST, Triage Assist and Remediation Assist. Checkmarx
A Linux Foundation study of 172 security practitioners found that 58% of teams ship AI generated code faster than security reviews can keep up, and that AI coding tools increased external package use for 60% of respondents. Microsoft FORGE Lab reports several soundness issues in Rust's standard libraries, including cases where safe Rust can trigger undefined behavior, found with its Hyenas agentic vulnerability scanner. Cloudsmith announced an AI Agent Skills registry at KubeCon that lets enterprises store, scan and control the skills that AI coding agents use. Linux Foundation study · Rust soundness · Cloudsmith
Agent governance and platforms
Uber described the security framework for its AI agents, with an agent registry, personas and intents, tool and MCP validation, runtime guardrails and continuous red teaming. Volume 35 of the Thoughtworks Technology Radar recommends stronger controls for AI agents, including sandboxing, threat modeling, access limits, human oversight and cost management. GitLab 19.5 made its MCP server generally available for tools such as Claude Code, Cursor and Codex, and added controls for agent permissions, session recording, rate limits and audit events. Uber · Technology Radar · GitLab 19.5
Anthropic documented enterprise managed authentication for Claude connectors, so administrators authorize apps centrally and users connect automatically when they first use an enabled app. Databricks MLflow now accepts native OpenTelemetry trace exports from the Claude Platform, including Claude Code and the Claude Agent SDK, and can store the traces in Unity Catalog. Dataiku introduced Headless, which lets developers create and manage Dataiku pipelines, models, dashboards and agents from coding tools such as Claude Code, Codex and Cursor. Claude connectors · MLflow traces · Dataiku Headless
Redis added six agent capabilities to its Iris context engine, with Intent Router, Agent Playbooks, Agent Memory and Context Gateway generally available and Lens and Agent Mesh in beta. Conduktor launched free AI agents in its Console that monitor Apache Kafka environments, investigate issues and recommend fixes for human approval. Google introduced GKE Workspaces, a self hosted development platform on Kubernetes with JupyterLab or VS Code, GPUs and TPUs, and isolated sandboxes for coding agents. Redis Iris · Conduktor · GKE Workspaces
Software modernization and testing
Orion Innovation made its OI NEBULA agentic software development platform generally available worldwide, covering legacy modernisation, AI coding assistant oversight, security fixes and deployment. Blitzy will work with Microsoft Frontier Company on modernizing large enterprise software systems, with integration into Microsoft Foundry, Azure infrastructure services and GitHub Copilot. Xebia is integrating CAST's software intelligence into Xebia Ace so enterprises can map application estates and plan modernization with agentic AI. OI NEBULA · Blitzy and Microsoft · Xebia and CAST
IgniteTech says its AI system finds root causes, creates and tests fixes, deploys updates and notifies customers, starting with Khoros Aurora AI and Iris AI under human controls. Vodafone plans to expand Witbe's AI testing framework across its eight country Vodafone TV footprint, from more than 120 continuous test cases to over 300 within 12 months. ANS Group reports that LogicMonitor's Edwin AI cut its alert noise by 83% and average resolution times by 44% since its August 2026 deployment. DataAgent released DataAgent Lab and an Apache 2.0 detection pack as open source, a local environment that uses K8sGPT and Kubernetes to build autonomous SRE workflows without access to production clusters. IgniteTech · Vodafone and Witbe · ANS and Edwin AI · DataAgent Lab
MCP connectors
McGraw Hill and Anthropic let US K-12 teachers at participating districts use licensed McGraw Hill curriculum in Claude for Teachers through the Model Context Protocol. Fitch Solutions made Fitch Ratings, CreditSights and BMI Intelligence content available in Perplexity through its Fitch Nexus MCP connector, with source attribution and compliance logging. Cashmere launched Meridian with Cambridge, Johns Hopkins and Princeton university presses, giving authorized students and researchers access to licensed books and journals through AI tools. Claude for Teachers · Fitch Nexus · Meridian
Awardco made its MCP connector available for Claude, Microsoft Copilot, Gemini and Glean, to draft and send employee recognitions and look up recognition data, with existing permissions and human confirmation for actions. Awardco
Wolters Kluwer launched an MCP server for TyMetrix 360°, so users can query live legal management data and manage matters, invoices and tasks through AI assistants within existing permissions. TaxDome released a client MCP that lets taxpayers ask Claude, ChatGPT or other assistants about their documents, invoices and tax return status at no extra cost. Cashfree Payments launched Checkout for Agents, a WebMCP based layer where AI shopping agents read orders and select payment methods while customers still authorise the payment. TyMetrix 360° · TaxDome · Cashfree
HtAG Analytics opened API and MCP access to Australian sold commercial property data, starting at A$0.075 per row with the first 25 rows free. LenDb.ai is a free database of 224,703 US loan officers with a free API and MCP server, managed day to day by an AI CEO under a human board. STARII TECHNOLOGY added Koozee MCP to its apparel visual production platform, so supported AI agents can request batch image generation. HtAG Analytics · LenDb.ai · Koozee
GPUs and frameworks
NVIDIA CUDA 13.5 moves Vera Rubin support to general availability for enterprise production and adds GPU scheduling controls and batched memory copies in CUDA Graphs. PyTorch 2.15 replaces its NCCL backend with a design that uses one communicator per process group, supports GPU replacement and adds a flight recorder for diagnosing failures. Innatera opened a Developer Portal and made Pulsar neuromorphic evaluation hardware available through DigiKey in the European Union. CUDA 13.5 · PyTorch 2.15 · Innatera
South Korea's ETRI released TANGO, a browser based no code tool that automates AI model training, optimization and deployment, as open source. TANGO
Also
Texas Instruments added agentic AI tools to its CCStudio environment that generate code, configure devices and build, flash and debug applications for its microcontrollers, wireless and radar chips; TI says they are available now. Cavli Wireless shows Cortex AI, a firmware assistant that generates and compiles code and flashes devices, at electronica 2026 in Munich, and Tuya added Silicon Labs SiWx917 support with AI assisted coding tools to its TuyaOpen framework. TI CCStudio · Cavli Cortex AI · TuyaOpen
Cybersecurity company Axonius expects 66 employees to leave, including 46 in Israel, and CEO Joe Diamond said its engineering organisation is being reshaped around an AI driven approach to software development, Calcalist reports. Axonius