Daniel Iser

Educators & course creators 👋 Online education is evolving fast. I’m running a 5-minute survey to understand what’s actually happening on the ground. If you’ve ever: • launched a course • taught online • built learning experiences Your input matters. You’ll also get the 2026 State of Online Education Report when it’s released. Take the survey → lifterlms.com/state-of-onlin…

Should add, it would work for Codex or any MCP enabled Agent really. So this is Hooks for all.

Hey @theo 👋 Built agent-hooks for Claude Code - inject prompts AND trigger actions at specific workflow stages. session.start → inject context spec.drafted → call bash scripts, pipe to other chains batch.completed → async notifications Fire events, listeners respond with: • Template content (injected into agent) • Shell commands (sync/async execution) Event-driven, priority-ordered, zero coupling. Thoughts on this for composable AI workflows? github.com/danieliser/age…

That is exactly what I built this for, does planning by multi-phase panel reviews, executes with fan-out DAG patterned parallel runs. github.com/danieliser/str… But as for attention spanning of a real human, the second you have more than 3 running at a time, agents have downtime as you struggle to keep up with their speed of implementation.

@jeffrey_way Don’t you think you could, in four paragraphs, describe a body of work that could be broken into 100 individual steps, many of which could be performed 10 at a time, in parallel?

I'm pretty convinced that very very few developers are actually doing these super-sophisticated workflows that involve a dozen agents working concurrently in their own respective worktrees. Twitter/X would have us believe the opposite is true, weirdly.

@jeffrey_way But your 100% correct. If your reviewing even half what the agents are producing, you can't really scale beyond 2-4 chats/projects at any given time. Our attention spans aren't capable of keeping all of them moving at speed.

@jeffrey_way I use 3-4 chats at once, but each of those could be using 3-10 sub agents. Kick off a strategize session (on ultra) with the commands/skills below and you will 100% see fan-out agent patterns. Same for the execute plan and similar commands there. github.com/danieliser/str…

The "more details" I promised on #WebMCP for #WordPress. #AI How should AI agents safely interact with your WordPress site? We're figuring it out in the open. Give it a spin and let me know what breaks: reddit.com/r/Wordpress/co…

🔥 Fun fact — wppopupmaker.com may be the first #WordPress site running #WebMCP in production. Free plugin, works today. Your site can be next. More details soon! youtu.be/7A34ZNz2bMM

Hey. There's a lot going on. Ch-ch-changes! Join me and @mattmedeiros as we talk about it all. We'll put it all out there, dread, excitement, imposter syndrome… a dash of hope? It's a @TheWPMinute 💗 @WPProductTalk Collab! LIVE. Tomorrow. 10am Eastern youtube.com/watch?v=wymfD4…

Maybe we missed something, maybe our tests weren't thourough enough. Either way its up for review. We'd love to find we made a mistake and were one simple fix from 100% scores.

Our conclusion for @triggerdotdev users: If docs fit in context, paste them inline. The engineering effort to build compressed indexes, MCP servers, or skills may not be worth the quality loss. @gao_jude - would love your take on this.

🧵 @vercel claims AGENTS.md with compressed docs beats skills. We tested this for @triggerdotdev and found something they missed: Full inline docs beat EVERYTHING - including their compressed approach. cc @gao_jude