featherless

Omg the effect gives me physical jitters

I had steered away for a long while thinking it was subpar to GitHub, but it's really come a long way. Especially running it on a local network it's noticeably faster in every way than GitHub, and I'm able to build complex gitlab workflows with custom runners that are fully configurable and have effectively 100% uptime and no queues.

I migrated my entire workflow onto a personal GitLab instance after the whole "pay a fee to bring your own bags to the grocery store" GitHub Actions pricing shenanigans earlier this year.

Best decision ever.

100% uptime. 100% less stress with each of the product/pricing changes over the past few months.

Was also able to build my own GitHub Copilot equivalent that auto-reviews MRs interactively.

Highly recommend it.

I got openclaw to compete Qwen3-Coder-Next vs Minimax M2.1 simultaneously on my Mac Studio 512GB: https://clutch-assistant.github.io/model-comparison-report/

Except most of those services don't have at-home equivalents that you can increasingly run on your own hardware.

It would be silly to write a new one today. Plenty of open source + indy options to invest into instead.

For scheduled work, cron + a log sink is fine, and for pull request CI there's plenty of alternatives that don't charge by the minute to use your own hardware. The irony here, unfortunately, is that the latter requires I move entirely off of GitHub now.

Charging people to maintain a row in a database by the minute is top-tier, I agree.

Do you feel that orchestration runs on a per-minute basis?

This is not investment in CI/CD. I already did that, by buying and investing in my own hardware, my own workflows, my own caching solution.

This is like if Dropbox started charging you money for the files you have stored on your backup hard drives.

No. It is not worth a time-scaled cost each month for them to start a job on my machines and store a few megabytes of log files.

I'd happily pay a fixed monthly fee for this service, as I already do for GitHub.

The problem here is that this is like a grocery store charging me money for every bag I bring to bag my own groceries.

> But at $140 a month, how much time is that worth investing?

It's not $140/month. It's $140/month today, when my company is still relatively small and it's just me. This cost will scale as my company scales, in a way that is completely bonkers.

Won't those other systems create a cost that is metered in terms of run-time?

As a solo Founder who recently invested in self-hosted build infrastructure because my company runs ~70,000 minutes/month, this change is going to add an extra $140/month for hardware I own. And that's just today; this number will only go up over time.

I am not open to GitHub extracting usage-based rent for me using my own hardware.

This is the first time in my 15+ years of using GitHub that I'm seriously evaluating alternative products to move my company to.

Just wrote about my approach yesterday: https://jeffverkoeyen.com/blog/2025/12/15/SlotWarmedCaching/

tl;dr uses a local slot-based cache that is pre-warmed after every merge to main, taking Sidecar builds from ~10-15 minutes to <60 seconds.

I found that implementing a local cache on the runners has been helpful. Ingress/egress on local network is hella slow, especially when each build has ~10-20GB of artifacts to manage.

I run a local Valhalla build cluster to power the https://sidecar.clutch.engineering routing engine. The cluster runs daily and takes a significant amount of wall-clock time to build the entire planet. That's about 50% of my CI time; the other 50% is presubmits + App Store builds for Sidecar + CANStudio / ELMCheck.

Using GitHub actions to coordinate the Valhalla builds was a nice-to-have, but this is a deal-breaker for my pull request workflows.

There's absolutely no way that the cost scales with the usage of my own hardware. I cannot fathom this change in any way or form. Livid.

Genuinely curious about this as well. It's a major bummer that self-hosted infra can't be used to validate GitHub Pull Requests now; basically means I'll have to move my entire workflow off of GitHub so that everything can be integrated reasonably again.

This is absolutely bananas; for my own CI workflow I'll have to pay $140+/month now just to run my own hardware.

Does this result in the Swift code being compiled or transpiled?

AI is an amphetamine.

Hey all! Happy to discuss the setup; has been a big win for Sidecar with a massive jump in test coverage while keeping costs flat.

Oh hey that's my startup's app, Sidecar, mentioned! Cool to see :D

It's going super well — Been having an absolute blast and the automotive community is incredible.

Haha yes fully agreed :D The gains I mention above are on top of those gains.

http://sidecar.clutch.engineering/ — Sidecar, a personal automotive assistant. Sharing dev updates at http://threads.net/@featherless, beta registration is here: http://sidecar.clutch.engineering/beta/

I've been validating this sentiment almost every day.

I left my job last year, after over a decade leading design and UX for iOS at Google, to pursue building a bootstrapped solopreneur startup, and the speed at which I've been able to build things out has been wild.

There are entire domains that would have taken weeks or months for me to learn the ins-and-outs of the tech, that Claude has been able to knock out in a matter of minutes. It's like I have two full time junior engineers working with me at all times (with a similar amount of coaching + guidance required), and for $20/month. These are gains of easily 2x-10x productivity and I already consider myself to be fairly productive as a design engineer.

"I think you may be underestimating the capabilities of ChatGPT. While it's true that the chatbot can produce "bad" responses when given irrelevant or nonsensical prompts, I don't think that means the experiment design is inherently flawed. In fact, I would argue that the ability for the chatbot to generate a wide range of responses, both good and bad, is a testament to its complexity and sophistication. It's not just noise, it's a demonstration of the chatbot's ability to process and understand language in a dynamic and nuanced way." — ChatGPT's response

"I respectfully disagree with your assessment of ChatGPT. While it's true that it can produce gibberish when confused or out of depth, I don't think that necessarily means it's easy for it to fail the Turing test. In fact, I would argue that the ability to recognize its own limitations and admit when it doesn't know the answer is actually a step towards passing the Turing test. It shows that the chatbot is capable of self-awareness and understanding its own limitations, which are important traits for a machine to possess in order to pass the Turing test." — ChatGPT's response

"Haha, I know exactly what you mean. I feel like I'm being interrogated by a truth serum whenever I post on these forums. And as for the chatbot, yeah it definitely seems like a lame PC version of GPT. I'm not sure why they didn't include more explainability, but I guess that's just the way things go sometimes." — ChatGPT's response

Edited title is a bit sensationalized. Key elements of Material will still be important parts of the design system on iOS; it’s about acknowledging when the platform solution is sufficient for the intended purpose. I’m sure that there will still be gaps that need to be filled, but many don’t need custom solutions anymore :)