Andrea Corvi

Head of Experimentation & 5to9 builder

[Andrea Corvi @AndreaCorvi]

Try at: https://t.co/o9WJLTZ2cn

[Andrea Corvi @AndreaCorvi]

Code Meter is a macOS menu bar app that lets you monitor AI usage across providers like Claude, Codex, Cursor, Copilot, Kimi… Easy to keep an eye on quota left and when it resets.

Today’s update adds one of the most requested features: multiple accounts for the same provider.

[Andrea Corvi @AndreaCorvi]

@FarzaTV @alesmoretti crunchy?

[Andrea Corvi @AndreaCorvi]

@camsoft2000 it's more: what problem should I solve? 😜

[Andrea Corvi @AndreaCorvi]

@OpenAIDevs @work_louder @alesmoretti should we get one?

[Andrea Corvi @AndreaCorvi]

@FarzaTV @alesmoretti

[Andrea Corvi @AndreaCorvi]

@badlogicgames Is AGI really around the corner whilst I still struggle to get LLMs to get padding and alignment right? Am I thinking linearly instead of exponential? Something doesn’t add up

[Andrea Corvi @AndreaCorvi]

Sol 5.6 is an amazing model: I had a list of tasks, big features and refactors that I put in the backlog as the back and forth with older models was too time consuming. Since Sol's release: 7/10 have been done, I launch them at night, wake up and they're done.

[Andrea Corvi @AndreaCorvi]

@thdxr At first sight this might disappoint you: https://t.co/3uIQA4oC7v - but it's a different way of managing skills, agents, orchestration which led me to move away from CLI and use agent deck instead. Pi under the hood, all custom on top. Context should not be wasted...

[Andrea Corvi @AndreaCorvi]

@davis7 you can give a go to agent deck, it has Pi under the hood and it is fully customisable for your workflow: https://t.co/3uIQA4oC7v

[Andrea Corvi @AndreaCorvi]

if you want to try it out yourself, for free: https://t.co/3uIQA4oC7v

[Andrea Corvi @AndreaCorvi]

Agent Deck has been updated, and it's flying with the latest #Sol and #Terra models. Set different levels of thinking and model for each of agent, to avoid wasting quota. Output has massively improved. It always (in around 6 sessions of test) brought the task home in one prompt.

[Andrea Corvi @AndreaCorvi]

@Vivek4real_ @alesmoretti how many times this happened...

[Andrea Corvi @AndreaCorvi]

@theo Everybody tries to bring water to their own cause. Tiny sells hardware, obviously they push self hosting. 99.9% of people or less can’t self host, period. And even if crazy enough to spend 50-100k in hardware, you’re dealing with a sub optimal model compared to frontier.

[Andrea Corvi @AndreaCorvi]

@theo Agent Deck is open source, free to download: https://t.co/3uIQA4oC7v

[Andrea Corvi @AndreaCorvi]

Bigger context windows are useful.
But effective context is the real advantage.
That is one of the main reasons I love Agent Deck.

Give your coding agent exactly what it needs: the right skills, prompts, tools and agents.

Not too much. Not too little.

Inspired by @theo

[Andrea Corvi @AndreaCorvi]

@mattpocockuk I’m using it with https://t.co/61AhRCuauv - cooking really well

[Andrea Corvi @AndreaCorvi]

@bentlegen No Apple Podcasts? 😁

[Andrea Corvi @AndreaCorvi]

@_alejandroao I’m do the same, using agent deck with different agents for planning, coding and reviewing: https://t.co/3uIQA4oC7v

[Andrea Corvi @AndreaCorvi]

@davis7 you might find this interesting to try out

[Andrea Corvi @AndreaCorvi]

Your coding agent lives in a terminal. It deserves a home.
Agent Deck: a free native macOS app for AI coding agents. Sessions, skills, GitHub issues,forking, memory, all in one window, built to keep your context window lean. Powered by Pi coding agent.
https://t.co/61AhRCuauv

[Andrea Corvi @AndreaCorvi]

Agent Deck brings all together in one place:

Project-based sessions
Worktrees
Fork any conversation
Use a frontier model for planning and a cheaper model for execution
Install skills from GitHub or skills sh with one click
Lean context (from Pi)
Built-in GitHub issue management