AI Coding Assistants

Today's Stories

1. 1

   AI video market thrives after Sora stumbles; studios, marketers profit

   Hollywood studios and startups are deploying AI video tools at scale. Promise (backed by Google, Disney, and Silicon Valley investors) overlays AI-generated backdrops in real time during filming; Netflix plans to use AI in 300 of its 1,000 titles by 2026. Higgsfield, founded by former Snap executive Alex Mashrabov, raised $400 million at a $5.4 billion valuation, with annualized revenue jumping from $20 million to $700 million in a year. Inception Point AI manages over 100 AI personas across more than 5,000 active podcasts. Hybrid productions combining human actors and AI are estimated to be 20 to 50 percent cheaper than conventional films, and AI-assisted animated documentaries save 30 to 40 percent. This cost advantage could reduce filmmakers' dependence on major studios. Marketing platforms now let brands like Dollar Shave Club produce multiple videos per day instead of single campaign videos. However, resistance persists: directors including Christopher Nolan and Guillermo del Toro remain critical of generative AI.

   Higgsfield's customer mix has shifted dramatically—business customers grew from less than a quarter of revenue in January to a majority now. Inception Point AI operates profitably on as few as 20 listeners per podcast episode. Spotify plans to label AI personas with an 'AI Persona' tag. This contrasts sharply with Sora's trajectory: OpenAI's initial 2024 demos generated hype, but the late-2024 release fell short of expectations; Sora 2, launched in fall 2025 as a standalone app, was eventually discontinued, and a billion-dollar Disney deal fell through.

2. 2

   VS Code 1.133 adds pinned scroll, HTML auto-reload, Claude–Copilot mix

   Microsoft released Visual Studio Code 1.133, introducing pinned scroll (to keep prompts visible while coding), auto-reload for local HTML file previews, and the ability to use Claude and Copilot AI assistants together in the same session. Developers working with AI coding tools can now mix Claude and Copilot in one session rather than switching between them, and the pinned scroll feature ensures important context (like prompts or instructions) stays visible while scrolling through code.

   These features are now available in the official release; the mixed Claude–Copilot support expands flexibility for teams using multiple AI providers.

3. 3

   GitHub Copilot canvases make agent workflows visible and steerable

   GitHub has introduced canvases in the Copilot app, a feature that creates a durable, shared workspace where developers and AI agents can collaborate on code tasks. Two example canvases are now available: Java Modernization Studio (which cost approximately 3,000 AI credits to build) and Site Studio (which cost approximately 2,000 AI credits), both accessible through awesome-copilot for anyone to use or adapt. Current chat-based AI workflows become difficult to track as agents move faster than humans can review—important context like plans, decisions, and validation points get buried in scrolling history. Canvases solve this by making workflow state explicit and persistent, letting teams see what stage they are in, what decisions were made, and what needs approval, without constantly replaying context. This reduces repeated prompting, context loss, and rework, making recurring work more efficient and governable.

   Developers using Copilot agents can start by picking one repeated workflow and building a minimal canvas using the /create-canvas command, then contribute it back to awesome-copilot if it helps their team. The canvases are available now in awesome-copilot.

4. 4

   Snowflake launches MCP server and CLI for AI agents to access observability data

   Snowflake announced general availability of a redesigned Observe MCP (model context protocol) server and a new Observe CLI with full parity to the MCP. The new MCP server removes a custom LLM intermediary layer, allowing agents to connect directly to Observe's APIs and access the same capabilities available in the UI, including APM service maps, OpenTelemetry data collection setup, and active alert listing. The CLI provides programmatic access to Observe's full platform from the command line and works in agent environments like Claude Code as well as interactive terminal sessions. Observability platforms must now serve both humans and AI agents, not just people reading dashboards. Engineers are building custom alert-triage agents that automatically query Observe when alerts fire, and copilots that assist during incident investigations. By removing the LLM intermediary, the new architecture cuts latency and operating costs for every agent query, while giving agents direct visibility into Observe's data structure so they can determine which data sets to query and write efficient observability queries.

   Both the MCP server and CLI ship with prebuilt skills—structured workflows for common observability tasks like investigating incidents, tracing failures, validating changes, finding n+1 issues, and detecting outliers. These skills work immediately after configuration without additional setup. General availability is rolling out now across all clusters; eu-2 and ca-1 regions will follow shortly.