Hum

I launched Heyhum 11 days ago with basically zero expectations.

It was just this thing I built because I was tired of AI coding tools that lock you in and treat your data like it’s theirs. I figured if it helped me, maybe it would help a few other people too.

This morning I looked at the numbers and… people are actually *using* it. Not just landing on the homepage and bouncing. Opening the chat, trying the free tier, triggering events, adding the PWA. Almost 2,000 unique visitors in under two weeks, with zero real marketing.

I keep waiting for the angry bug reports or “this is broken” messages, but they haven’t come. It’s been strangely quiet in the best way.

The biggest thing I’ve learned so far? Shipping fast while your brain is screaming “it’s not perfect yet” is terrifying… but it turns out that “good enough” is sometimes enough for people to give it a real shot. My OCD has been throwing a fit the whole time, but the data is showing me that the thing I made is already useful to someone.

Still feels surreal.

If you’re building in public and feeling like nobody’s noticing, this is your reminder that sometimes they are, they’re just quietly trying it instead of shouting about it.

https://heyhum.net

Would love to hear how your own early traction has felt. The surprises, the doubts, the little wins that keep you going.

Hi everyone,
​I’ve spent the last few weeks frustrated with the "subscription trap" of modern AI coding tools. I wanted something that stayed out of my way, respected my privacy, and let me bring my own API keys so I wasn't locked into a $20/month commitment just to use the models I already pay for.


​So, I decided to build it. I call it Hum (Heyhum.net).


​I’m a stay-at-home mom of two toddlers, so I don't have the luxury of months of quiet development time. I spent the last 7 days in a state of hyper-focus, coding primarily on my phone and in the gaps of my daily routine, to get a stable version live. Nap time with a baby asleep on my chest works out to be the pretty good window for focus.


​The Core Philosophy:
​True BYOK: No middleware taking a cut. You provide the keys; Hum provides the interface.
​Privacy-First: It’s stateless. Your code doesn't live on my server.
​Surgical Edits: Built for "Vibe Coding" = making precise, intentional changes rather than letting an agent rewrite your whole file.
​Compare & Merge: Run two models (like Claude and GPT-4o) side-by-side with different prompts to see which one handles a logic problem better.


​The Reality Check:
The site has been live for a week. I’ve already seen a 10× spike in traffic and survived my first massive wave of automated bot scans (the "welcome to the internet" ritual).


​I'm currently on a total spending freeze after investing $400 into the initial assets and infrastructure. My goal isn't to build a VC-backed unicorn; I just want to build a tool that solves a real pain point for developers who value control and precision.
​I’d love to get some feedback from this community. Does the "Bring Your Own Key" model solve a pain point for you, or do you prefer the convenience of an all-in-one subscription?

I built Hum because I wanted to use my own model (especially Claude) in an environment that’s actually designed for coding, without giving up control or simplicity.

Most tools either lock you into one provider or feel like black boxes that rewrite too much.

Hum is a browser-based AI coding workspace that:

• Uses your own API key (Anthropic, OpenRouter, etc.)

• Supports local models via Ollama

• Makes surgical edits instead of full rewrites when possible

• Reads files on demand instead of loading everything into context

• Shows diffs before anything is applied