Explained What shipped, explained: every real release, decoded for the people who use it
The lane
A release, a changelog entry, a pricing change — made easy to understand: what it is in plain words, why it matters, who can use it and how, as technical as the topic demands and no more. Not news; the reference.
Prefer this site in Google
Mark omidsaffari.com as preferred and Google lifts it in Top Stories, AI Overviews and AI Mode for you.
Latest articlesWhat shipped, what broke, and what it really cost — from running production AI systems.
Cloudflare Shows Which Worker Slowed a Customer Request
Follow a slow request across Cloudflare Workers and Durable Objects, find the slow call, and check tracing costs before rollout.Sep 17, 2026Explained
Vercel Hobby Can Remove Old Previews Before 30 Days
Vercel changed Hobby deployment retention. Check which previews and rollback targets survive, when cleanup starts, and what to preserve.Sep 17, 2026Explained
Cloudflare Can Stop AI Spend Landing on the Wrong Bill
Cloudflare AI Gateway can require your provider credentials. Learn when missing keys stop a request and which charges stay separate.
Sep 17, 2026Explained
Cloudflare Lets Python Apps Reuse Existing Databases
Python Workers can connect through Hyperdrive. Check what that changes for an existing database, app architecture and hosting bill.Sep 16, 2026Explained
Gemini 3.8 Live Keeps Callers Talking During Lookups
Gemini 3.8 Live runs tools during voice calls. See what changes for booking flows, customer updates and the cost of a completed task.Sep 16, 2026Explained
Cloudflare Limits What a Client's Deploy Agent Can Change
Cloudflare adds access controls for individual Workers. See how to separate debugging, code review and deployment rights across client projects.Sep 15, 2026Explained
Claude Code Stops One Install From Widening the Whole Job
Claude Code can approve network hosts for one command at a time. See what that changes for dependency installs and unattended build jobs.Sep 15, 2026Explained
Vercel AI SDK Can Move Agent Spend to Existing Plans
Vercel AI SDK can use supported agent subscriptions. Check which credentials win, which allowance pays, and what sandbox costs remain.Sep 15, 2026Explained
Cloudflare Browser Run Keeps Client Jobs on Approved Hosts
Limit client browser jobs to approved hosts, budget for required CDNs, and let reviewers watch through read-only Live View.Sep 14, 2026Explained
GPT-Live-1 Changes the Budget for AI Phone Calls
Understand GPT-Live-1 phone-agent costs: the voice layer, backend reasoning, telephony, and the interruption handling worth testing.Sep 14, 2026ExplainedThe other lanes.Pick the one you came for.
143 pieces
AI
Models, agents, coding assistants, image, video, voice136 piecesBuild
Ship with AI when you're not an engineer: vibe-coding, app/SaaS builders, automation, MCP, just-enough infra135 piecesDesign
For designers, brand owners, creative directors138 piecesGrowth
AI for marketing, SEO/GEO, content engines, and the operator-business work of turning audience into revenue6 piecesPlaybooks
AI for your niche: the workflows that actually move, tool by tool, priced0 piecesLab
The agent journey, entry by entry: what shipped, what broke, what was decided, in the ventures I run36–45 / 45
Gemini agentic video understanding, explained: when 88% fewer tokens is real
Gemini can now inspect only the parts of a long video it needs. Here is how agentic mode works, what it costs, and when static mode still wins.Sep 2, 2026Explained
Claude Fable 5.1, explained: cheaper context, stricter rules
Claude Fable 5.1 cuts cache reads to $0.25 per million tokens, but its API changes can break tool calls and edited agent histories.Sep 2, 2026Explained
Vercel AWS PrivateLink, explained: what Pro teams get
Vercel now gives Pro and Enterprise teams a private route to AWS services. Here is the setup, pricing, limits, and decision rule.Sep 1, 2026Explained
Codex CLI 0.152.0, explained: what actually changed
Codex CLI 0.152.0 adds per-tool MCP output limits, longer app-server shell timeouts, Vim draft search, and clearer recovery states.Sep 1, 2026Explained
Vercel's fx AI SDK harness adapter, explained
Vercel added fx to HarnessAgent. Here is how the ACP adapter works, what it costs, and the five limits to know before production.Sep 1, 2026Explained
Cloudflare Adaptive Intelligence, explained
Cloudflare’s bot model now retrains continuously. Here’s what is live, who gets it, how to enable it, and which promised parts still aren’t shipping.Aug 31, 2026Explained
Cloudflare AI Search with GLM-5.3 Flash, explained
Cloudflare AI Search now generates answers with GLM-5.3 Flash. Here's how the pipeline works, what it costs, and when to switch.Aug 30, 2026Explained
Grok Build's Vercel AI SDK adapter, explained
Vercel put Grok Build behind AI SDK's HarnessAgent. Here is the setup, the auth path, the ACP limits, and who should use it.Aug 16, 2026Explained
Vercel's AI SDK ACP harness adapter, explained
Vercel's new @ai-sdk/harness-acp package connects ACP-compatible coding harnesses to HarnessAgent. Here's when to use it and where it breaks.Aug 16, 2026Explained
Nano Banana, Explained: Google's Gemini Image Models (Pro, 2, and the Original)
Nano Banana is Google's Gemini image model line. What Pro, 2, and the original each do, real per-image pricing, how to use them, and which to pick.Jun 19, 2026ExplainedNewsletter
One letter, every Sunday.Working systems, not hot takes.
Weekly. No spam. Unsubscribe anytime.