Julio Barros

I'm opening beta access to Guided Surveys — an AI interviewer that replaces static forms with real voice or text conversations. Participants tap a link and talk. You get analyzed insights in minutes. Looking for beta testers: guidedsurveys.com

Agentic Cafe meeting March 25 online at 5pm Eastern. Philip Faulconer - "Current thoughts on Claude best practices" then open up for questions and other topics. After that, April 13th @ 4pm Eastern and Wednesday April 29 @ 5pm Eastern. Come join us: agenticcafe.com

Drowning in .md files from all my AI work, so I built a simple Mac markdown viewer — fast, clean, no editing. iPhone/iPad coming soon. Check it out on TestFlight testflight.apple.com/join/SMyvPYgR

Notes from our latest Agentic AI meetup: - agents processing order emails - AI-assisted coding workflows - LLM PR reviews - multi-agent system challenges Lots of practical experimentation happening right now. agenticcafe.com/articles/2026-…

You don't always open the app anymore. You invoke it. New post on the dissolving shape of software: skill-based apps, agent-native apps, workflow apps, and meta-apps. E-String.com/articles/what-…

We had great Agentic Cafe meeting on Wednesday. Don Shin helped kick everything off with a short talk on his agentic processes. I wrote up some notes at agenticcafe.com/articles/2026-…

Happy Twin Peaks Day to all who celebrate. youtube.com/watch?v=nRK9uE…

Next Agentic Cafe is Wednesday Feb 25, 7–8pm ET. Small group, practitioners only. On the table: OpenCode experiences, orchestration, open discussion. agenticcafe.com

Most arguments are convincing because they only represent one perspective. I built a system where AI agent, each representing a different stakeholder, analyze, debate, and stress-test the argument from every angle. Here's how it works: e-string.com/articles/agent…

A while a go I had an iPad app that some people really liked, iCardSort. Which is a digital card sorting app for organizing ideas, concepts, etc. Its coming back and a TestFlight version is available at testflight.apple.com/join/sMT7MqYt LMK if you have any issues or questions.

Our first Agentic Cafe meeting was yesterday and it was great. Really impressed with the group and the quality of the discussions. I wrote up some notes here: agenticcafe.com/articles/our-f…

Xcode Agents, Cursor, Codex, Antigravity, Claude Code Turns out programming gets replaced with juggling in the AI era.

The new Claude Code tasks (along with AskUserQuestion tool, and skills etc) means that you can create your own custom workflows. I like seeing the tasks come and go and their status, etc so wrote a Mac app CCTaskViewer to do that. Its on testflight testflight.apple.com/join/wZeBH8Es

@chris_karani "In the cloud" with third party services or do you think Apple's servers may end up being enough?

It’s been so much fun building agents in Swift. Understanding on device constraints have been the best part. I believe the future is more hybrid, you’ll have some processing on device and some in the cloud.

After 40 years of writing code, I'm barely writing code anymore. Not because I've stopped building—because agents do the typing now. The job is shifting from syntax to specs. From coding to designing intent. Wrote up what I'm seeing and where to start e-string.com/articles/shift…

Just shipped SwiftAgents. Its Langchain but built in swift github.com/christopherkar… #swift #langchain #aiagents @langchain @SwiftLang

I’ve been using claude code extensively lately and they’ve provided passes for a free week if anyone is interested   claude.ai/referral/edjxm… FWIW

I've been working on a voice assistant app and have been learning a lot about the topic. This is a summary of my related notes. e-string.com/articles/voice… btw checkout the app at getaudioinput.com

In a world where coding agents are as good if not better than the best human coders, you’re not going to have much of a edge if your primarily skill today is cranking out code for projects that someone else came up with. Instead, you can differentiate yourself by having great ideas for what to build and being able to actually ship those ideas. The best way to get better at both of these is to build a lot solo. The more you come up with ideas for things to build yourself, the more you build them, and the more you put them out there in the world for others to use, the better your taste and end-to-end execution skills will become, and the more valuable you will be in the near future when superhuman coding agents are ubiquitous and your coding skills are commoditized.

I've gotten questions about various agent protocols and acronyms so I wrote up the more common ones such as MCP, A2A, AP2, etc, at e-string.com/articles/ai-pr… Let me know if you have any questions or if I missed something critical.