pcannon

Founding Engineer @AquinF03

[pcannon @pcannon775]

I'm very happy on how much Aquin has evolved, started as a dream hobby, and now it's a dream come true

From creating mid-software to one of the best ideas that we've made in a while

I remember that Aquin started as a online website utility (Calculator, Currency converter, ...), but then we pivoted to AI when we saw the opportunity to be something real

We were still an online website at that time, where you could chat with a lot of chat-bots, customization inside of your workspace

Later, we've thought; What if we make something truly unique? Something that hasn't been completely done before, hence, Aquin Lucid came out, an AI browser that you could have customization, an intuitive UI layout with some banger features. Sadly, that product was discontinued when we thought that we could do something better, something that you wouldn't need to replace your default browser to have more features

Then Aquin's floating assistant came out, where you didn't need to replace your favorite browser to just have features that you wanted; AI, MCPs, Google connections and even talking with your browser tabs

Suddenly, we've thought; What if training an AI was accessible to anyone? Even with those who don't have supported GPUs or limited hardware? So we made research, we made progress, even a page in our website to get to know our view on LLMs and making the AI training possible (https://t.co/fyxmQq8NI2)

So then, in the middle of the development, something great happened; @Ashf03 got accepted to @theresidency and we were constantly developing, non-stop, we got help by so many great people at The Residency, I've never imagined being this far, though, there's still lots more to come, even things that I cannot simply predict

We launched Aquin yesterday, we made incredible progress throughout the years even if are small steps to being a bigger dream

Check out the trailer here below, one of the best so far

👉 https://t.co/oc0vmznhVU

[Ash @Ashf03]

Aquin is out. vibe code your LLM in 2 mins.

[Dev Hunt @devhunt_]

👋 Friends!

⏰Just 12 hours left to vote for your favorite developer tool at DevHunt.

21 amazing devtools are competing for the "Tool of The Week" title 🥇

Finalists :
→ https://t.co/ufbdpmd2tS
→ Stacktree by @stevy_smith
→ Policy Template Generator by @55i5
→ https://t.co/GH0tB0XU2R
→ Aquin by @Ashf03

Your vote is important, so don't miss out!
→ https://t.co/54NIS27Tal

[pcannon @pcannon775]

@zuhaitz_dev I wanted this for a long time, is it in GitHub or somewhere else?

[pcannon @pcannon775]

@AquinF03 What next, `sudo apt install aquin`?

[pcannon @pcannon775]

@AquinF03 PS: The background image remembers me about the Lotus III game from the Amiga computer

[pcannon @pcannon775]

@AquinF03 i made this for u

[pcannon @pcannon775]

Preparing for @AquinF03 demo at 10PM IST, quite nervous

Let's hope everything goes well!

[Aquin Labs @AquinF03]

Starting a little research fellowship!

- 2 individual researchers grouped in duo working on linguistic features with quantization methods

- 2 AEIA UC Santa cruz lab AI research assistants working on some neuro stuff with embedding models

we'll be demoing and giving compute today 10PM IST, join to hangout and for our first public demo!

Event: https://t.co/d2Vxfa8rWh

Server: https://t.co/9FaEGmxdp4

[pcannon @pcannon775]

@zuhaitz_dev @thdxr Greatest food of all time 🔥

[pcannon @pcannon775]

@Ashf03 @AquinF03 Imagine just how much you can save with Aquin without needing to re-train everything again and again until it's decent - this can be a live saver for many companies, and especially with startups!

[pcannon @pcannon775]

You won't believe how much compute gets wasted on exploratory fine-tuning. So at @AquinF03 we built something that predicts what helps or hurts before training: bad samples, dead layers, useful concepts. You get a final checkpoint without running a training loop.

[Ash @Ashf03]

for the first time you can simulate fine-tuning before committing any compute

with @AquinF03, see exactly which features strengthen or suppress, which samples hurt generalization, which layers go dead!

[pcannon @pcannon775]

I don't even sleep with a pillow lol

[pcannon @pcannon775]

Code it in Python, extend the DevKit's functionality, apply it to any of your software, and by any i mean ANY. Package, push and voila!

[pcannon @pcannon775]

Aquin DevKit is a small SDK and CLI where you can track and analyze ML training runs, and specifically LLM Fine-tuning where it can record metrics localliy, save a final model checkpoint and upload results to Aquin's website to visualize the data neatly

🧵↴

[pcannon @pcannon775]

Aquin DevKit is basically a modular inspector which lets you upload any training runs and inspect, not just how well a model trained, but what changed inside the model in low-level way

[pcannon @pcannon775]

I escaped the matrix for ~4 months, finally!

[pcannon @pcannon775]

I remember getting stung by a wasp, and i asked Siri what should i do, Siri just searched up the internet, it took one full minute and guess what happened... *Sorry i can't understand you*

[Dogan Ural @doganuraldesign]

Hey Siri, set a timer for 10 minutes

[Dev Hunt @devhunt_]

⇨ Aquin by @Ashf03 is a local-first platform for building private LLMs, fine-tuning models, and creating RAG pipelines without infrastructure complexity.