Agent infrastructure matured while the models became subscriptions
Vercel Sandbox, Waymo rides at SFO and Prism, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.
Vercel
Vercel Sandbox
DEPLOYif you run coding agents in production
WHAT SHIPPEDVercel Sandbox is generally available as an execution layer for agents, with on-demand Linux microVMs, sub-second starts, snapshots, and Active CPU pricing. Vercel processes over 2.7 million deployments per day on the same Hive compute platform.
THE CATCHYou pay only for active CPU time, not idle time, and snapshots let agents restore a known state instead of recloning repos and reinstalling dependencies. The CLI and SDK are open source.
IF YOU RUNagent workflows that spin up dozens of sandboxes per task and need isolation for untrusted code
SOURCEVercel, Vercel Sandbox is now generally available
Waymo
Waymo rides at SFO
WATCHif you travel through SFO
WHAT SHIPPEDWaymo began offering fully autonomous rides to and from San Francisco International Airport starting January 29, 2026, with pickups and dropoffs at the SFO Rental Car Center. Access starts with a select number of riders and expands over the coming months.
THE CATCHInitial service is limited to the Rental Car Center via AirTrain, not the terminals, and only a select number of riders get access first. Waymo's Bay Area service area covers 260 square miles.
IF YOU RUNa Bay Area commute or airport run and want a driverless option
SOURCEWaymo, Clear skies and autonomous Waymo rides at SFO
OpenAI
Prism
DEPLOYif you write or review scientific papers
WHAT SHIPPEDOpenAI introduced Prism, a free AI-native workspace for scientific writing and collaboration powered by GPT-5.2. It is available today to anyone with a ChatGPT personal account, with unlimited projects and collaborators.
THE CATCHPrism is built on Crixet, a cloud-based LaTeX platform OpenAI acquired, and GPT-5.2 works within the document with access to equations, references, and structure. More powerful AI features will move to paid ChatGPT plans over time.
IF YOU RUNa research group juggling LaTeX, citations, and co-author revisions across institutions
SOURCEOpenAI, Introducing Prism
The Deploy Log
Models and APIs
| WATCHif you pay for Google AI Ultra Project Genie. Google DeepMind rolled out Project Genie to Google AI Ultra subscribers in the U.S. aged 18 and over, a prototype web app powered by Genie 3 that generates interactive worlds in real time from text and image prompts. Google DeepMind |
| DEPLOYif you code in the terminal daily Mistral Vibe 2.0. Mistral released Mistral Vibe 2.0, a terminal-native coding agent powered by the Devstral 2 model family, with custom subagents, multi-choice clarifications, slash-command skills, and unified agent modes on Le Chat Pro and Team plans. Mistral |
| DEPLOYif you build on ServiceNow ServiceNow chooses Claude. ServiceNow chose Claude as the default model for its ServiceNow Build Agent and as a preferred model across the ServiceNow AI Platform, and is rolling out Claude and Claude Code to its global workforce of more than 29,000 employees. Anthropic |
Tools and products
| DEPLOYif you triage anomalies in Slack Vercel Agent in Slack. Vercel Agent investigations now appear directly in Slack alert messages as threaded responses, and 10 investigations are included at no additional cost for Observability Plus subscribers. Vercel |
| DEPLOYif you run Claude Code against Vercel Vercel API CLI command. vercel@50.5.1 adds a new api command giving direct access to the full suite of Vercel APIs from the terminal, with agents like Claude Code inheriting user access permissions automatically. Vercel |
| DEPLOYif you build interactive frontend interfaces Kimi K2.5 on AI Gateway. Kimi K2.5 is now accessible via Vercel AI Gateway with no other provider accounts required, with advanced coding abilities especially for frontend code quality and design expressiveness. Vercel |
| WATCHif you need models that self-select tools Qwen 3 Max Thinking. Qwen 3 Max Thinking is now accessible via Vercel AI Gateway with no other provider accounts required, autonomously selecting built-in search, memory, and code interpreter tools during conversations. Vercel |
| DEPLOYif you compare models by live performance Live model performance metrics. Vercel AI Gateway now displays throughput and latency metrics across hundreds of models, updated every hour, with P50 and P95 latency and throughput available in the model list, model detail pages, and REST API. Vercel |
| DEPLOYif you manage skills for coding agents Skills 1.1.1. skills@1.1.1 adds interactive skill discovery with npx skills find, a meta find-skills skill for AI agents, non-interactive mode, support for 27 coding agents, and is now fully open source. Vercel |
Left out
- Claude Code Max via Gateway: a figure in the draft could not be found on the publisher's own page
- AssistLoop on Marketplace: it did not carry a number that moved
- cubic on Agents Marketplace: it did not carry a number that moved
- Trinity Large Preview: it did not carry a number that moved
- Tag-based cache invalidation: it did not carry a number that moved
- Skew Protection for prebuilt: it did not carry a number that moved
- Daggr: it did not carry a number that moved
- GOV.UK AI assistant: it did not carry a number that moved
The pattern
The infrastructure for running agents became a product in January 2026, while the models that drive them became subscriptions. Vercel Sandbox is generally available as an execution layer with on-demand Linux microVMs, sub-second starts, snapshots, and Active CPU pricing, and Vercel processes over 2.7 million deployments per day on the same Hive compute platform. Vercel Agent investigations now appear directly in Slack alert messages as threaded responses, and 10 investigations are included at no additional cost for Observability Plus subscribers. vercel@50.5.1 adds a new api command giving direct access to the full suite of Vercel APIs from the terminal, with agents like Claude Code inheriting user access permissions automatically. skills@1.1.1 adds interactive skill discovery with npx skills find, a meta find-skills skill for AI agents, non-interactive mode, support for 27 coding agents, and is now fully open source.
On the model side, OpenAI introduced Prism, a free AI-native workspace for scientific writing and collaboration powered by GPT-5.2, available today to anyone with a ChatGPT personal account, with unlimited projects and collaborators. Google DeepMind rolled out Project Genie to Google AI Ultra subscribers in the U.S. aged 18 and over, a prototype web app powered by Genie 3 that generates interactive worlds in real time from text and image prompts. Mistral released Mistral Vibe 2.0, a terminal-native coding agent powered by the Devstral 2 model family, with custom subagents, multi-choice clarifications, slash-command skills, and unified agent modes on Le Chat Pro and Team plans. ServiceNow chose Claude as the default model for its ServiceNow Build Agent and as a preferred model across the ServiceNow AI Platform, and is rolling out Claude and Claude Code to its global workforce of more than 29,000 employees. Kimi K2.5 and Qwen 3 Max Thinking are now accessible via Vercel AI Gateway with no other provider accounts required, and Vercel AI Gateway now displays throughput and latency metrics across hundreds of models, updated every hour, with P50 and P95 latency and throughput available in the model list, model detail pages, and REST API. Waymo began offering fully autonomous rides to and from San Francisco International Airport starting January 29, 2026, with pickups and dropoffs at the SFO Rental Car Center, and its Bay Area service area covers 260 square miles.