ToolsHow Codex is widening who can build at loveholidaysloveholidays has adopted Codex to let non-engineering teams prototype and deploy software changes, increasing AI-assisted code edits from 7% to 79% in a year and boosting deployments 73% without growing engineering headcount.4 min read
ToolsRun SDK: secure eval with QuickJS for running dynamic Code ModeThe developers of Run SDK introduce the secure eval feature, which allows running dynamic Code Mode code in a lightweight, QuickJS-based secure environment.1 min read
ToolsMario, the creator of Pi, recommends a video about memory; Claude Code's poor memory comes upMario (the creator of Pi) recommends a video about memory management that analyzes the limits of coding tools' memory; the video highlights Claude Code's particularly poor memory performance.1 min read
ToolsNVIDIA Dynamo’s shadow engine speeds LLM worker recovery from minutes to secondsNVIDIA Dynamo’s preview feature “shadow engine recovery” keeps a fully initialized standby engine on the same GPUs as an active engine and uses a GPU Memory Service (GMS) to share weight allocations without duplicating them in HBM.6 min read
ToolsVercel Connect generally available: secure service access for agentsVercel announced that Vercel Connect is now generally available (GA), making it easier for agents to securely connect to services and data.1 min read
ToolsDesign systems as the control layer for AI-generated user interfacesAI tools can rapidly produce working frontend code, but that speed risks fragmenting a product’s UI when each generated feature introduces its own patterns and decisions.9 min read
ToolsAI Gateway can now call webhooks after long-running generationAI Gateway has introduced the ability to call webhooks after long-running generation processes, which can reduce compute costs and increase reliability because the Gateway takes on a larger processing burden.1 min read
ToolsClaude and Claude Cowork have received a shared memory controlled by the userClaude and Claude Cowork now use a shared memory into which the user can decide what is placed.1 min read
Toolsexa_search integrated into Gateways: fast, zero-configuration search for agentsThe exa_search tool added to gateway services allows agents to use fast, zero-configuration Exa search; no credentials, account, or separate connection are required, and billing is handled via…1 min read
ToolsPerplexity and Nvidia bring Portable Computer to run agentic AI fully on local machinesPerplexity and Nvidia today launched Portable Computer, a local-first version of Perplexity's agentic Computer platform that runs models, files and orchestration on users' own hardware with no token billing for local work.8 min read
ToolsWispr Flow turns spoken prompts into structured text for faster AI inputWispr Flow records spoken prompts and converts them into structured, polished text intended for use with various AI tools.2 min read
ToolsAnthropic unifies Claude and Claude Cowork memories and adds user-editable memory storageAnthropic announced that it is merging the memory systems used by Claude chat and Claude Cowork so the assistant will retain information learned in one context when used in another.3 min read
