Hec Perez

Still, this feels like a glimpse of how software development workflows might look soon.

Also: it’s great to see OpenAI leaning on Elixir for orchestrating agents.

One thing I miss is the ability to attach to the running Codex sessions.

Ideally I’d like to watch (or even modify) what the agent is doing before it finishes.

You get some visibility through Linear’s issue plan/logs, but it still feels limited.

The GitHub account Symphony uses can create PRs but cannot merge to main.

Technically Symphony can merge to main automatically if you want.

I’m not trusting it with that yet.

For code review, I initially just ran git diff inside each issue’s directory.

Later I configured Symphony’s WORKFLOW.md to automatically open a PR in GitHub.

Then I open SSH tunnels from the server to my laptop for those ports.

That way I can access the running instances locally, test them, and review the changes before approving anything.

One workflow I tried:

When an issue moves to Human Review, Symphony spins up a server for that issue (each on a different port).

So every issue has its own running environment.

I’m running it on a Hetzner server, letting it operate with full freedom.

Symphony can handle multiple issues in parallel, creating a dedicated directory for each one.

OpenAI launched an Elixir app this week: Symphony.

It’s an orchestrator for coding agents that start working on issues when you move them into the Todo column in Linear.

I’ve been experimenting with it over the past few days and it’s a pretty interesting direction for agent workflows.

Who's going to the ElixirConf EU Málaga? Early Bird tickets end in less than a week!

Drop a 👋 below if you're going, I'd love to connect in person.

I beta tested this and it’s amazing.

I asked Claude Code: find all html and accessibility errors of YouCongress via Rocket Validator’s MCP tool and fix them.

After a few runs, all green. ✅

The tool call just failed

Add MCP Server (#174) · youcongress/youcongress@0cc808c

And yes, it's based on the mcp-session-id github.com/youcongress/...

Add MCP Server (#174) · youcongress/youcongress@0cc808c

By the way, I fixed the link in the Livebook of my workshop to the YouCongress source code where I forward the MCP events: github.com/youcongress/...

If a multi-node MCP server uses StreamableHTTP/SSE, when a node receives a tool call, it needs the session of the http connection.

I solved it by forwarding to the original node, but redis or a global ETS table like you did is cleaner 🙌

But what does all this have to do with OpenRouter?

Here you have bsky.app/profile/hecp...

Here you have. I hope you like it! bsky.app/profile/hecp...

Here you have bsky.app/profile/hecp...

Elixir Madrid workshop materials: LLMs, Agents & MCP in Elixir apps.

Livebook examples → notes.club/hectorperez/...

#ElixirLang

Building MCP servers for ChatGPT and API integrations Learn how to build MCP servers for use with ChatGPT connectors, deep research, or API integrations.

ChatGPT Pro and Plus too. I don’t have a pro plan anymore there but it seems it should still work. By the way, I remember their UX when using MCP worse platform.openai.com/docs/mcp

It’s great it stopped snowing and you came! Actually, in the morning I feared that we may need to cancel the whole event. Good to see you there!

This Wednesday, I’ll be teaching a workshop on LLMs, Agents, and MCP in Elixir.

We are fully booked! 🚫🪑

However, if you want me to send you the Livebook notebook with all the content afterwards, just drop a comment or DM me.

#ElixirLang

Eustaquia: A smiling plant powered by Erlang on embedded hardware | Icia Carro, Thu, Feb 12, 2026, 7:00 PM | Meetup Imagine a plant that greets you with a smile when it’s well watered and slowly turns sad when it’s thirsty. (Please enter your full name on https://tally.so/r/WOExxj to ac

Second event of the year in Elixir Madrid: a smiling plant powered by Erlang on embedded hardware by Icia Carro, an Erlang/OTP project on embedded boards + pizza & drinks. www.meetup.com/madrid-elixi...

Why the AGI race may be bigger near-term existential risk than misalignment A Fermi-paradox metaphor for coordination failure

The biggest near-term risk from AI isn’t “evil superintelligence.”

It’s a coordination failure.

I expand on why this matters using the Fermi Paradox as a metaphor:

👉 open.substack.com/pub/networks...

Just tried asking Claude (desktop) to create a simple Livebook tool for me, and it worked!

Once done, I can click "Open in Livebook" and it opens directly in my Livebook desktop 🎉

Try with "create a livebook to..."

#ElixirLang

Empiezo el año impartiendo este taller de IA en el meetup de Elixir Madrid.
@jaimeiniesta.com gracias por compartir 🙌 ¡Nos vemos allí!

Human testimony in the age of AI Why public decisions need sourced quotes, not just AI answers

Most AIs compress disagreement into one answer.
Governance needs the opposite: sourced quotes, stance-labeled (For / Abstain / Against), linked to primary sources.
open.substack.com/pub/networks...

The Perpetual Technological Cage Unless superintelligence is developed under a global consensus, the risks will be shared by all, but the upside won't. This is why I signed the superintelligence statement.

I signed the "superintelligence statement". Yet, beyond "flourish or doom", there’s a quieter risk:

If the US/China build it now, they'll enforce a "non-proliferation" to keep it that way.

That’s how a Perpetual Technological Cage begins for the rest of us:
networksocieties.com/p/the-perpet...

We’ve removed AI-generated perspectives. All expert and public-figure stances on YouCongress now come from verifiable, sourced quotes—so you can delegate with confidence.