Give an AI agent Redis memory and verify recall across sessions
Save a real authorized user fact in Redis Agent Memory, inspect automatic extraction, then verify a fresh LangGraph agent received and used the scoped memory.
The BIG CHANGE archive
News, analysis and practical guides.
Explore our reporting on how AI and robotics change work, business and the world.
The latest change
849 stories
How to make a checked real estate flyer and inquiry form with AI
Turn approved property facts and photos into checked listing copy and a flyer, connect a live inquiry form, test every link, then reply within recorded permission.
How to verify an AI-assisted Oracle-to-PostgreSQL migration
Use AWS's scoped AI schema conversion with separate DMS row migration, matched-checkpoint validation and a tested reverse-CDC cutover plan. A documentation-based guide for authorized DBA teams.
Measure MCP tool context without losing the answer
Build a pinned local MCP server and Claude host, compare five context controls against a broad baseline, and check answers before token totals.
Run one bounded AI agent overnight and inspect its morning digest
A documented n8n recipe for one read-only night work order: approved notes, a bounded agent, saved state and failure alerts, then human review of the morning draft.
Make a checked infographic, process diagram and visual explainer with AI
Use Canva and Lucid AI to draft three distinct visuals, then check infographic figures, process branches and explainer panels against approved source records before exporting.
Build and check a no-code two-agent support triage in n8n
A documented n8n recipe for a restricted support form, two tool-using AI workers, shared run state, fail-closed checks and a human-reviewed draft. Readers verify actual production executions and worker overlap.
How to run and review a GPT-5.5 multi-file change in Codex
GPT-5.5 remains available to eligible API-key Codex projects. Set a real two-file coding task, verify the active model, then inspect tests and the complete patch before acceptance.
Build social carousels, quote cards and a 30-day visual calendar with AI
Use approved brand facts and Adobe Express visuals to make checked Instagram image slides, a distinct LinkedIn PDF document, quote cards and 30 dated account-specific posts. Verify permissions, exports, scheduling and final status.
Build a Claude brand kit, landing page and launch video
Start with real brand facts, review a logo concept and reusable rules, then check an exported Claude landing page on a phone and play the finished Motion MP4.
Run parallel Claude Code research subagents and verify their findings
Set up two read-only Claude Code research subagents, check their actual task and model state, then reconcile cited findings, failures and usage in the main session.
Build an editable investor deck with Claude from approved evidence
Turn approved venture, market and competitor evidence into a sourced Claude Slides deck, then reopen its PowerPoint or Google Slides export for fact, rights and editability checks.
Move a Claude web project through GitHub and checked Vercel deployments
Review a runnable Claude Code project or usable Claude Design export, push approved source to GitHub, then verify Vercel's first production release, branch preview and later production update against their commits and access rules.
Build and check a seven-role local and cloud AI model portfolio
Configure five local and two permission-gated cloud AI task routes, then check exact model identity, hardware fit, output quality, latency, privacy and cost before accepting each route.
Turn a mobile app spec into a checked Claude Design prototype
Use an approved app spec to build every required Claude Design screen and interaction, check the prototype on a mobile device, then verify the reviewed export or share.
Mistral launches Voxtral 3 speech models and voice-agent SDK
Mistral's Voxtral 3 links separate speech models, reasoning and tools in a voice-agent SDK. Its docs show a browser build path, prices and phone limits.
Microsoft says Volkswagen has given Copilot access to over 330,000 staff
Microsoft reports more than 330,000 Volkswagen staff have Copilot access, over 300,000 monthly Copilot Chat users and 26,000 active agents. The announcement provides no measured productivity result.
Meta pauses Muse sharing after researchers find chat pages through passive DNS
Meta paused Muse conversation sharing after Claroc found thousands of readable share hostnames in passive DNS. Claroc's October 20 recheck found 40 older pages no longer opened, while their titles remained in DNS history.