Why do we need another AI coding assistant?
There’s no hardware focused coding assistant that our team actually loves. We see a gap, we pounce (at least until the Anthropic overlords deem us worthy).
Anyway, this post is an insider take on how and why we plan to bridge that gap using Hydron: our take on what a hardware coding assistant should be.
First some history - the last full-fledged hardware projects I did were way back in undergrad, playing around with Arduino, ESP-32 and 8051 boards. It was fun, but it was always cumbersome. You had a pile of docs to read through, and the coding itself felt unnatural, like the tools were fighting you. I've been out of hands-on hardware for a while since, but I still follow every word in the daily product calls (given I do my daily homework).
Fast forward to now - here’s a shameless plug: Now that Hydron is available, I am back to developing projects that run on my current ESP32 S3 board using a bunch of cool new sensors I picked up from SP Road. These are not full blown embedded engineering projects with STM32, NXP, Jetson Orin level boards but still a great place for me to quickly test where Hydron stands. HW testing using Hydron is a must-do activity now for everyone at H2Loop, mandated by Sai, our CEO, because how else are we going to develop an assistant that speaks hardware without first understanding its language.
Everyone around me is excited to go back to the realm of Hardware. As we all might’ve heard in one podcast or another, software had its day in the sun and that day is now long over. Or at least that’s what all the tech gurus tell me and maybe we believe them.
Hardware is the next big MOAT. Physical AI, the next big frontier.
To fully realise Physical AI and see it proliferate among the masses though, we need to break all the barriers surrounding true hardware development. Make it so modular, so easy and so accessible that you go from prompt to a working robot in a matter of days.
That’s the vision with Hydron (long, long term. Don’t come at me yet)
Everyone has to start somewhere, and we at H2LooP believe building Hydron out whilst testing it on actual hardware is a great way to start. It sits at this beautiful intersection of Software, Hardware and AI and if the vision is executed well, it has the potential to be something huge.
While we were developing, there were a lot of internal and external discussions on what Hydron should look like and what it should do. Talking to embedded engineers helped. Actually creating a hardware lab in-house helped even more. Hence our ever expanding embedded hardware team (we even have a founding engineer role open - check out our careers page).
We launched a stable Hydron ready for the public to use a few weeks back. It already helps embedded engineers across verticals write firmware code and debug it much faster. But the vision, that Physical AI vision, that’s where the meat is. We are taking many steps toward it and we’re very happy to have started this journey in earnest.
Numero uno - if you've ever brought up a board, you know reading and re-reading datasheets is the single biggest time suck. So we built a hardware context engine that automates that away and delegates the referencing and spec-citing to the Hydron agent.
Numero dos - bringing up a board is never just one datasheet. You're toggling between 2-3 platforms, a dozen tabs and a pile of documents just to get the board talking, and it only gets worse where safety critical applications are involved: more certification docs, more tools, more platforms. How are we to realise the true boundaries of Physical AI if we're still switching tabs at the end of it? So we're building toward one surface, closing the HW-SW loop and containing it to a single place. That's the final boss.
That leads me to our roadmap: everything agentic, everything from within your native surface. From debugging tools to certification artifacts to vertical specific personas, we are working on a whole bunch of ideas, integrations and partnerships to effectively close the loop. Things have started but we expect it to take time for it all to synergise and compound. How long till we get there? Damned if I know but hey it’s the journey that matters :P
And yes, finally, to answer the original question of why do we need another AI coding assistant? I’d tell you to stop thinking of Hydron as a coding assistant. Start thinking of it as a holistic hardware development ecosystem waiting to burst at the seams!.
So come one, come all. Build with Hydron today. Finish your long shelved dusty hardware hobby projects (or) your unfinished degree projects (or) even just use it for Python code if that’s your thing. Tell us where we fail and how we could do better.
– Ajith
feat. a bit of AI (~20%)
Check us out: https://www.hydron.sh/
Book a Hydron demo today: https://calendar.app.google/paohM7mUrJA9vzeS7