Manuel Cecchetto

[Manuel Cecchetto @manuelcecchett]

https://t.co/hjEBA9YmmY

[Manuel Cecchetto @manuelcecchett]

I keep thinking about how aggressively we’re optimizing AI agent workflows for software/tech, while so many higher-impact fields still feel “behind”.

I’d love to help researchers, doctors, teachers, and people doing work that directly improves lives use these tools better.

[Manuel Cecchetto @manuelcecchett]

@nicdunz I need to work I don’t need a friend.

[Manuel Cecchetto @manuelcecchett]

Farei una cosa molto semplice.

Se hai più di 55 anni non devi cacare il cazzo sul progresso.

[Repubblica @repubblica]

IA, Mattarella contro le big tech: “Pochissimi privati travolgono ogni regola, l'Ue agisca” https://t.co/vH7GQZINrx

[Manuel Cecchetto @manuelcecchett]

@Ruffino_Lorenzo Smonta anche un po’ di quella che è la dietrologia dei provider che ci stanno facendo un favore a venderci l’inferenza a così poco

[Manuel Cecchetto @manuelcecchett]

GLM-5.2 is a full time internal yapper. The thinking summary is literally 100-120 paragraphs of yapping.

[Manuel Cecchetto @manuelcecchett]

@thorstenball Are you sure? = You’re right to point that out.
Shouldn’t we …? = You’re absolutely correct, we should.

[Manuel Cecchetto @manuelcecchett]

@ThePrimeagen The way:

Using https://t.co/iLCHdi7t3z, hermes, openclaw, /loop, /goal, /btw, /gstack-toilet-paper, /grill-with-my-docs, /cuckCode, MCPs, self improving skills, subagents, local glm-5.2 turbokv cache, llama.cpp, 12 mac studio m5 ultra max 512gb unified memory.

[Manuel Cecchetto @manuelcecchett]

@nicdunz We’re on a big echo chamber, people have no idea. Basically 0.05% of people

[Manuel Cecchetto @manuelcecchett]

@thsottiaux I know what’s coming!

Bug fixes and small improvements!

[Manuel Cecchetto @manuelcecchett]

Wanted to try the GLM-5.2 coding subscription, can I use it on pi?

[Manuel Cecchetto @manuelcecchett]

@nicdunz Finally i needed my daily bugfixes and small improvements

[Manuel Cecchetto @manuelcecchett]

@DataChaz New and MCP in the same tweet. Give your claude web search bro, he’s living in 2025

[Manuel Cecchetto @manuelcecchett]

@eurofounder Classic french

[Manuel Cecchetto @manuelcecchett]

@teortaxesTex If this is enough to become sota and smash all the benchmarks, let the lab-shuffle begin!

[Manuel Cecchetto @manuelcecchett]

@satyanadella Brother your company is the worst disgrace that the tech world has ever produced, even AI couldn’t improve the situation.

Just silence, that’s what we want from you.

[Manuel Cecchetto @manuelcecchett]

AI made this landing page.

Claude? Codex? no.

Ideogram4!

Good luck implementing it tho 🤣

[Manuel Cecchetto @manuelcecchett]

@thsottiaux Why removing the possibility to decide what compaction prompt you can run on the codex app/cli? Compaction is a big part of the harness loop i want control of

[Manuel Cecchetto @manuelcecchett]

We could release Rome 23 Open 2T tomorrow.

Should we? Is 🇮🇹 back?

[Chubby♨️ @kimmonismus]

Wait what? Rio 3.5 Open 397B, developed by IT company of Rio de Janeiro's city government is now SOTA open source and even outperforming Qwen 3.7?

What is happening today.

Never heard of them before.

[Manuel Cecchetto @manuelcecchett]

They cooked.

[Pavel Durov @durov]

We now support rich formatting for all chatbots.

Tables, nested lists, inline media, formulas, headers and more — right in Telegram messages.

🔨 Start building! Docs: https://t.co/zgzPOOUJF5

[Manuel Cecchetto @manuelcecchett]

@droidbuilds Design.

Which isn’t a national security threat I believe?