octember

Claude Tag seems awesome, but a single query can run over a dollar. I can't afford anthropic api rates, but i can afford a codex plan

earshot rides exe.dev on your personal plan, so for a small team it's effectively flat-rate. it sits in channels and chimes in when it can help. I've wanted this for a long time, and anthropic was kind enough to show us what it should be. I wrote a spec off it, so you can point your own coding agent at SPEC.md and regenerate it in whatever stack

The whole thing is one process and one sqlite file.

Happy to answer anything

Might be time to check gemma :)

Cool idea, but keyframes are not videos. Motion, object permanence, are not things Claude can infer from a set of images. Nice demo though!

I write a lot of code.

AI agents are rats in a maze: they reach for what they know. And unless you show them better, what they know is slop.

MOST software is garbage, and they'll happily imitate it.

You can't just let AI run loose if you give a damn about the craft. So we babysit. I constantly yell at them to stop writing terrible code. It works, but takes forever, and had become a big bottleneck in my work.

For code to be maintainable, it needs to be simple, boring, even "dumb". You shouldn't need a PhD to read production software.

So i built stupify - it reviews my agents against code I actually respect, and yells so i don't have to.

The quality of review is anecdotally much better than BugBot and CodeRabbit, for pennies on the dollar.

Give it a shot, curious what HN thinks.

https://github.com/Octember/stupify

Bevyl | Founding Engineer / AI Engineer | NYC / open to remote | https://bevyl.ai/ Bevyl is a seed-funded startup building AI-powered content creation tools for consumer brands. We're making it possible for brands to produce high-quality social media content at scale without agencies or massive creative teams. The other competitors in the market are laughably bad, and we have a chance to build something really powerful on the cutting edge of what's possible.

You'll work with me (CTO, ex-Airbnb/Loom/Hightouch) to build key parts of our video + AI tooling. Video is hard, and we need to move fast to win. We have real customers, a lot to build, and strong conviction in our space. This is genuinely a very fun project and I'm looking to build my core team.

Stack: TypeScript, React + Tanstack, PostgreSQL, Supabase, AI pipelines. Competitive salary and significant early-stage equity

You: Early startup experience, willingness to hustle, love for product, comfort with ambiguity & startup pace

noah@bevyl.ai

Bevyl | Founding Engineer / AI Engineer | NYC/open to remote | https://bevyl.ai/

Bevyl is a pre-seed startup (team of 2) building AI-powered content creation tools for consumer brands. We're making it possible for brands to produce high-quality social media content at scale without agencies or massive creative teams.

You'll work with me (CTO, ex-Airbnb/Loom/Hightouch) to build key parts of our video + AI tooling. Video is hard, and we need to move fast to win. We have real customers, a lot to build, and strong conviction in our space. This is genuinely a very fun project and I'm looking to build my core team.

Stack: TypeScript, React + Tanstack, PostgreSQL, Supabase, AI pipelines. Competitive salary and significant early-stage equity

You: Startup experience, willingness to hustle, love for product, comfort with ambiguity & startup pace

noah@bevyl.ai

Tech is saturated with SaaS apps. The days of a viral app store launch are gone. What remains are boring industries in need of innovation, but little appetite for new tools

SaaS can still work, but self-serve and solutions are more important than ever.

Let's say you've built the perfect SaaS: you've built the perfect tool for a customer and put it in front of them, excited to save them time/money/energy. And?

They don't want it! They don't want to learn your new fancy software.

But what do they want? Solutions: someone to do the job for them. That person doing the job could be leveraging SaaS on the backend, but these businesses have 0 appetite for doing the new work themselves.

AI enables solutions to scale. Operations, communication overhead, manual sales processes: this is all ripe for automation + AI. What was once a fixed linear cost scaling has become a more attractive curve.

I've been using both ChatGPT and Claude daily for months, and noticed many of my peers switching to Claude as their workhorse LLM. As a product engineer, I wanted to understand why.

My take: ChatGPT's advanced features (persistent memory, infinite customization) actually create UX friction for most users. Claude's opinionated defaults and consistent personality just work better out of the box.

This mirrors classic product lessons: the iPhone won not because it had more features than Android, but because it required less configuration. Feelings beat features in consumer tech.

Curious if others are seeing similar adoption patterns? And for those who prefer ChatGPT, what keeps you?