We've been shipping fast — Ghost, Gamma, HeyGen, Linear, HubSpot, and dozens more are now on IFTTT. None of it is about any single app. It's about the ones you already use finally working together, automatically, in the background. ifttt.com/explore

A year ago, building a new service integration for @IFTTT took our team 1–2 weeks. With AI, it's now 2–3 hours. That's the AI story I find most convincing — not agents replacing people, but small teams shipping like big ones.

GOAT

nice

The automations I love most are the ones I've forgotten I set up. New mention → Slack alert. Post scheduled in buffer → goes live a couple days later. Files change on my laptop → gets backed up later that evening. No app to open, no agent to prompt. It just runs, the same way every time. Quiet reliability is the whole point.

The AI software I'm most excited about isn't smarter answers. It's use case specific tools that actually know my context. Two people with the same apps connected shouldn't get the same results or even the same interface. It should all be personal. The real unlock is software that gets more useful the longer you use it — because it understands you better.

For years, "connect everything" had an asterisk: everything with an API. The interesting frontier now is the long tail that never had one — the small service, the old tool, the site with no integration. AI is starting to let automation reach those too. That's when "everything" gets real.

Every day I run across 5-10 "thoughtless acts" (@JaneFultonSuri) — the tiny inventions people make all day without noticing. Looping a bag over a doorknob. Wedging a book to prop a door. etc. We improvise our physical world constantly. The digital one should be just as easy to bend to fit our needs.

Google just launched Gemini Enterprise — agent orchestration and governance for big companies, right behind ChatGPT Work and Anthropic's enterprise push. Every big platform is racing to give IT departments controllable agents. Fair question: who's building that for the individual?

Biggest MCP (Model Context Protocol) update since launch ships this month. Connecting was never the hard part. Auth needs daily upkeep, each tool hogs your context window, and there's still no "if this"—no triggers, no subscriptions. Right idea. Still aimed at engineers.

X just shipped an official MCP server, joining GitHub, Slack, Stripe, and Notion. More of the world talking to AI is good. The "if this"—the trigger that starts the work or helps AI know when new data is available—is still missing. Also notable that usage of their MCP server requires a dev account/app. This isn't MCP for everyone. docs.x.com/tools/mcp

Grok. Perplexity. ChatGPT. Claude. Gemini. All on IFTTT now. Not for chatting—for actually doing things across 1,000+ apps and devices. ifttt.com/explore/whats-new-…

The interesting shift in AI right now isn't smarter answers. It's that the models are starting to act — send, post, buy, book. Which makes the boring question the big one: can you trust what it just did? Human understandable structure is what makes an agent safe to set and forget.

MCP's the right idea, but collectively aimed at the wrong audience. Easy to connect, impossible to maintain auth without daily/weekly upkeep. Missing triggers. Endless permission prompts. Eats up huge chunks of your context window. It won't really take off until it's treated as a protocol for everyone using AI to get work done, not just engineers.

current conditions

workshop inspiration

Grok just landed on IFTTT. Here's the thing about AI assistants: great at deciding what to do, useless at reliably doing it across your 200 other tools. That's the part we handle. Plug Grok into an applet and it actually acts — the same way, every time. ifttt.com/grok

AI agents hit 66% task success this year, up from 12% (Stanford AI Index). Everyone reads that as "structure is dead — just let the agent figure it out." I read it the opposite way. The more an agent does for you, the more you need to see and trust what it did. Structure is the feature.

The narrative going around: AI agents will make "if this then that" automation obsolete. Maybe someday. But we're not in the year of agents — we're in the decade of them. Until then, the tools people actually rely on pair agent smarts with structure they can see and trust.

@Kalshi + @Polymarket are here. 🌍 Set custom price threshold alerts, get notified the second a market resolves, and automatically log every move to @googlesheets , so your prediction game is always one step ahead. ifttt.com/kalshi | ifttt.com/polymarket