Joined June 2023
- Tweets 1,408
- Following 323
- Followers 77
- Likes 3,318
222 Photos and videos
Pinned Tweet
11 Mar 2025
Full Stack Dev | PHP, Laravel, Vue.js
Building solutions, some games, fixing bugs, and drinking coffee ☕
I share insights on coding, optimization, and dev life.
More about me: brunoserrate.serratedevs.com…
3
666
Ah yes, again, my country suprising me "positively" again
What a joke
3h
The Rio 3.5 model broke the internet this week. The plot twist? It’s essentially our open-source model, Nex N2 Pro, wearing a different hat.
🤯 We analyzed the weights, and the recipe is exact: Rio 3.5 ≈ 0.6 * Nex N2 Pro 0.4 * Qwen 3.5
It even literally introduces itself as "Nex N2 Pro" if you ask it without initial system prompt!
😂 We are flattered that the City of Rio used our work to achieve SOTA performance. Thanks for the ultimate benchmark validation.
🤝 But in the open-source world, attribution matters.
👇 Full mathematical proof & verify script in the first reply!
8
Jun 13
Like, what?
Brazilian AI model made by IT company in Rio that is better than qwen 3.7? Damn
Jun 13
Alibaba Qwen3.7 slowly fading into irrelevance at the frontier due to proprietary stance.
In it's place we have Minimax M3 and... *checks notes* Rio 3.5 397b, made by the municipal IT company of Rio de Janeiro's city government.
huggingface.co/prefeitura-ri…
1
86
Jun 13
Alritght, time to test it
Intelligence should be open, accessible, and ready to build with, empowering every developer, everywhere.
GLM-5.2 is now available to all GLM Coding Plan users, including Lite, Pro, Max, and Team plans.
docs.z.ai/devpack/latest-mod…
As our new flagship model, GLM-5.2 delivers powerful coding capabilities, usable 1M-context support, and continued strengths in long-horizon tasks.
API and Chatbot services will launch next week. The model will also be officially open-sourced next week under the MIT License.
The future of AI is open, and it belongs to the people.
36
Jun 12
Day 4 of the Post Your Mind logbook.
A lesson I learned while building: a good feature doesn't turn into dead weight later. Even though I still prefer the good old email password, a lot of people want to centralize everything in a single account. That's why, since launch, we've had social login with Google and GitHub (OAuth done right from day 1).
Options for users are always really important, I like them, why wouldn't they, right? And you? Do you prefer the control of a password or the ease of a single click via Google/GitHub?
1
29
Jun 12
But the feature I was genuinely most excited about is THREADS.
That insight that's too long for just one post, but too short to turn into a blog article?
With threads I can make linked posts, tell a full story or build on a topic slowly, keeping the thread alive.
Perfect for breaking down a thought into 3 or 4 connected posts. In my tests, that's exactly what I felt was missing.
1
37
Jun 12
Still on the same vibe: I centralized all AI providers and models in one dynamic spot. Switching providers, adding models, tweaking configs… all in one place.
That "invisible" refactor that completely changes the user experience. Less headache now and later.
What d'you think? Which feature do you use most or would love to see around here?
#PostYourMind #BuildInPublic
8
Jun 12
Today I had to go to an appointment and there is no way to take the notebook to the street like this.
I am glad that we have Hermes to do the tasks on the telegram for me.
Thank you to the entire Nous Research team for this amazing tool.
38
Jun 12
Has anyone figured out or knows a workaround to get Ollama Cloud usage? I really want to integrate it into my project, but it seems like there's a PR that's been open for ages and still no update.
19
Jun 11
Me and my inability to stick to building just one thing.
Ai Quota Tick is a simple program to track your AI quota. Yep, there are already several projects like this, but not the way I want, and not for Windows. By the way, it's open source and 100% local!
99% of the project was done with minimax-m3, 1% with GLM 5.1, and 100% with hermes.
Link to the repository in the comments below.
1
20
Jun 11
There's this really funny thing about building my products. i'm really not into communication, but it always ends up messing with it. building Post Your Mind and other conversational tools, it's actually quite interesting. so many challenges. controlling something not deterministic.
13
Jun 11
Day 3 of the logbook: I stopped handing out on a silver platter how many users I have.
The MVP was born with sequential IDs (1, 2, 3…). Great for testing. Terrible for putting into production. Anyone hitting the API instantly knows how many users you have, how fast they're growing, how many records exist total.
I migrated all PKs and FKs to random non-sequential IDs. Annoying? Yeah, but it's the kind of thing you'd rather do before launch.
#PostYourMind #BuildInPublic
1
1
33
Jun 11
While I was at it, I took the chance to turn the MVP into a real product.
I set up Stripe (credits subscription) and, most importantly: credit calculation is now based on the actual tokens each AI call uses.
None of that "charge 10 credits per generation and hope for the best."
The user pays exactly for what the AI spent.
1
1
32
Jun 11
MVP "ugly but works" turned into a product with real billing and secure IDs.
And did you migrate from sequential IDs to random ones?
And when it came to adding billing, did you go all at once or gradually?
Tell me how your switch went. I'm really curious.
1
10
Jun 10
Nice, now im gonna have to give my coding agent a sarcastic British accent in SOUL.MD and when he answers, generate with that voice. 'Your features ready, now let me grab a cup of tea for the review' LOL
Rlly awesome work
Jun 9
@Teknium hey 👋
i work on Gemini at Google
i added a few unique Gemini TTS features to Hermes to make the experience much more fun: expressive audio tags and “directors notes”.
this lets the hermes agent sound incredibly lifelike in its tts responses
check out my PR github.com/NousResearch/herm…
1
33
Jun 10
I think Dario's approach is really stupid. Like I heard once: He has a god complex, that's why he does what he does.
I don't blame an "entity" called Anthropic. It's Dario who has that. He keeps the model to himself and does doomsday marketing.
I even prefer Anthropic's models for work, but being stuck with this provider, like this at sky-high costs? No thanks.
What's crazy to me is that Fable is blocked from life sciences broadly, nerfed even if you get passed the classifiers and filter level blocks.
The whole point of AGI/ASI is to cure all diseases. Everything else is just nice to haves. But Anthropic wants to close off that path.
I think Anthropic might be the worst company on the planet.
24
Jun 10
Day 2 of the logbook. AI just really got into Post Your Mind. Non-negotiable requirement: It HAS to respect what I wrote. No "I'll just polish it a bit" and hand over a text that doesn't sound like me.
1
14
Jun 10
What I want is simple (and tedious to get right):
Grammar correction that fixes and STOPS.
Tone I set.
Translation that keeps the Portuguese English mix.
Platform as input (Twitter ≠ LinkedIn). All of this changes the text completely. The AI has to know where the post is gonna live.
1
19
Jun 10
And the core of the product? Full versioning.
Each version saved and marked:
"this here was me" × "this here was the AI".
Without that, the AI never learns what your real voice is.
Next posts: how this scales and what goes down when the AI starts to really know you.
What do you think? AI that amplifies your voice or that "improves" everything?
12