Marco Rodrigues

Today is Father's Day in Portugal 🇵🇹 2+ years of being 50% parenting, 50% freelancing. Exhausting. But I get to spend quality time with my daughter every single day. Something I won't be able to experience ever again. She grows fast, and I'm watching it in real-time ❤️

I thought AI would replace developers by 2030. I was wrong. It already started. I closed more Upwork contracts in 2025 than in 2024, and my yearly revenue increased by more than 2x. So I thought that growth would continue this year. But I was naive. AI is now doing 60% of what I used to do. Clients are using Claude, Codex, and other harnesses to handle data, create dashboards, build pipelines, and much more. On top of that, invitations on Upwork dropped significantly. So I found myself in a difficult situation during the first months of the year. I had two options: 1. Accept the daily LinkedIn invitations to work for companies. 2. Try something new. The first option seemed like the easiest path, but at the same time, I’m concerned about the longevity of programming positions. Even if I applied for one, there’s a good chance it could become obsolete within a year. Instead, I decided to learn more about “Harness Engineering”, in other words, learning how to orchestrate agents like Claude, Cursor, and open-source frameworks such as OpenClaw and Hermes Agent by @NousResearch. It took me two months, and things shifted. Now, instead of coding, I’m mostly vibe-coding, setting up agents across multiple fields, creating second brains, building workflows, debugging, and customizing agents. I have so much demand that I’m even thinking about hiring my girlfriend, who recently lost her job as a web developer (due to AI). If you’re a jobless developer in 2026, maybe it’s time to think about alternatives. New domains are emerging, and we need to ride the wave.

@ellev3n11 Improve code quality, without me having to tell it to organize the code in multiple scripts and classes every time 😅

what do people want to see in composer-next? any new particular capabilities?

@rezoundous It's good to build fast. Not to build well. But if you give it the right instructions, it can fix the mess (a bit).

Is Composer 2.5 really that good at coding? Anyone tried it yet?

@HDonda It's super fast and free!

the underrated part of this: platform docs ai usually gives you the snippet with the right version baked in. claude has to guess which release you're on, which method signature is current, which thing got deprecated last quarter. saves a half-hour of "wait is this still the current api".

📌 Tip of the day. If you're debugging something related to a platform, tool, etc. Don't ask Claude or ChatGPT. Go to the platform's docs page and Ask AI. Most companies have this nowadays, and there's no better place to debug than querying the actual Knowledge Base.

@AleBisquiz @Teknium For what? Coding?

@dadhalfdev @Teknium gemini 3.5 flash is awful

Anyone tried in Hermes yet? New OS king?

@hanrgb @SHL0MS @Teknium Hire me. I will help you. Don't get frustrated 🙂

@SHL0MS @Teknium Fair point, but it still doesn’t address the issue of losing focus across all these tools, where you end up just using the specific skills you need anyway. Also, dms.

Can someone give a concrete argument for why one should use Hermes Agent instead of building everything custom? I understand that the tools are widely available, but to me it feels like it introduces unnecessary complexity and confusion by forcing you to work within someone else’s architecture instead of building your own system from scratch, where you fully understand every part of it.

@spencermerrill @Teknium For agents what I want is speed and high agentic benchmarks. Coding is not so important for me tbh, because I tweak the scripts with Claude or Cursor if needed.

@dadhalfdev @Teknium I'm curious to see prices. From what I've been seeing thus far on agentic tasks, is that Gemini 3.5 Flash -- although fast -- is pricey.

If you're setting up a custom Hermes Agent for a client. One of the most important files is the "policy.yaml" in the tirith folder. Add all the rules to protect from attacks.

@Teknium This removes the "skill creation" when giving instructions for workflows?

Skill Bundles in Hermes Agent allow you to predefine a set of skills, then force load them all with one slash command. Makes workflows that rely on a set of skills super easy! Get access to them early with `hermes update`

There's something nobody tells you when you decide to be a freelancer. Your working hours need to be 100% productive. This is not something that happens inside a company. You're obliged to be there 9-5, but nobody checks your productivity levels beforehand.

As a developer. Is there any advantage to using the agent mode in @cursor_ai instead of the chat only? Is it only for pure vibe-coders, or am I missing something?

@Bl00d3clypse France is more Mediterranean than Portugal. There's nothing Mediterranean about Portugal 😅 the sea starts after the Gibraltar's strait, Spain.

I love how we always exclude France in exchange of Portugal Anyways hot take: mediterranean europeans arent just these 4

@yo_itsmatt @superteam Best season in Lisbon, enjoy!

gm from lisbon ☀️ we need a @superteam here

@theo Source? Working very well for me and insanely fast.

Oh my god it scored worse than Composer 2! Not even 2.5! And it cost 4x more to run!!! This might be the worst major lab model drop of all time. Llama 4 tier. Insane.

Everyone was criticizing Google Gemini 3.1 Pro for being slow. Just tried Gemini 3.5 Flash. It's insanely fast and smart!

Gemini 3.5 is available on OpenRouter. Let's see how fast it actually is!

@Google Already!

@Google When available?

Meet Gemini 3.5 Flash — our strongest agentic and coding model yet. It delivers frontier-level performance at 4x the speed of comparable frontier models — often at less than half the cost. Generally available, starting today. 🧵 #GoogleIO