I've seen some projects that use it and you open the PR page to be greeted by every PR having 3-20 comments but when you goto the actual PR, there's no one except the contributor with a bunch of Copilot feedback.
It gives a false message that the PR is resonating with folks and has real activity. I wonder if GitHub did this on purpose to make engagement seem higher than it really is.
I want to know how many real humans read my post, commented, shared etc.
Clankers can keep their own counts.
[0] https://github.blog/news-insights/company-news/bringing-more...
It seems that for "actions", the trailing twelve months availability is 98.67%.
Trailing 3 months is even worse :/
I have always found it as a pretty nice to have feature if I am already using GitHub. It’s far from perfect or robust but I can get a lot of use out of it with low to no friction.
It is time to setup local models. It is cheaper, and you already have a computer. Why keep it idle and pay someone else for their CPU?
Once it is cheaper, there will be more demand so it will no longer be cheaper. Buying now gets current prices (though demand is still fairly high).
There are more and more independent AI inference providers without VC backing that serve open weight models on a ~cost-plus basis that show that subsidies are not significant for AI inference.
E.g. a well-designed deployment (infrastructure-as-code) repository doesn’t need a frontier model to be understood well-enough to create a new job / service using sibling jobs / services as templates.
And this already saves me dozens of minutes per week, although it’s not a 2x multiplier in my efficiency.
They never really get tight very long: the various states are way too busy flooding the world with endless money printing to kick the can of the public debt always further.
Covid financial crash? We went to new highs. 2022 tech flash crash (Meta and Netflix did -75% for example): we then went to new highs.
The only way for governments who ever spend way more than they bring in taxpayer dollars is to de-valuate the currency.
So "financial markets getting tighter": probably won't last.
So you'll probably never see government customers allow that and neither will a lot of commercial customers.
I don't see the risk. If your code is easily AI generated you don't have a moat anyways. A Chinese competitor probably won't have as easy of a time as a US one of you operate in the US
Further, at a lot of companies, the risk has to be acceptable to shareholders and auditors. Perceived risk is often a more powerful motivator than actual risk.
lmao tell that to the artists, authors and foss contributors whose work has been cloned into the llm oracle
Also, how long do you think openai, Microsoft, Google, anthropic, etc could delay a lawsuit while you pay hundreds of thousands in legal retainers? 5 years? 10?
From the perspective of someone currently living in the EU... I'd say thats pretty much a wash (or even slightly tilted in China's favour) for folks outside the US
Maybe the government should nationalize GitHub at this point as it is absolutely critical for US infrastructure and MSFT has shown to be a terrible steward for the public.
If you want to make your repos public, you could use cgit and the like.
> Last month, we shared how GitHub Copilot code review runs on […] GitHub Actions using GitHub-hosted runners.
But also - enterprise accounts already have budget assigned to github actions, and this allows them to start billing right away without having to actually get (or allow) businesses to evaluate the return of having copilot do code reviews.
So seems like it's a mix of immediate incentives and long term architecture. I don't like it, though. If I were an enterprise my first response would be to turn it off.
Hang on, I read this as copilot reviews with bill both actions minutes and AI credits. Did I miss something?
Though actually the more I think about it, I think this change actually does make more sense. In the case of the AI running on GitHub side, that does feel pretty equivalent to CI minutes. I would hope that the number of minutes they bill for is pretty minimal though, since the vast majority of that will be I/O waiting on the agent to return
Done that way it obfuscates cost of the code review and I think that's on purpose
That’s why.
Do they, though? I don't know a single person who uses GitHub who actually likes it. It's far more often something like "it's fine, but I miss (GitLab|Gerrit)" or "I stopped using it for personal stuff and moved to (Codeberg|GitLab)."
The brand recognition among non-technical folks is really the strongest selling point in my eyes. And that's irrelevant to ~95% of software development.
Stopped my recurring subscription at the end of last year when it started spinning up actions for review. Which as a side effect doubled the time (or so) to do a review. Whereas before that I would open a PR, wait at most a minute or two and the review was already done.
I’m blowing through my 1000 mins in days.
Thinking to either pool some free tiers or figure something out with spot instances.
Also is it just me or is CI/CD tooling still sort of rough all around.
Hetzner has cheap VPS that I host my CI on. It costs like $10/month.
Pick the cheapest region, since CI runners location doesn’t matter much.
But I think the issue is that my situation (solo dev, mono repo) is just not right for a dedicated instance.
With only 1-2 runners, the pipeline is slow (low parallelism) and resource constrained. And at least 50% of the time its idle (I'm not working/sleeping).
I guess what I'm really looking for is for some kind of aggressive autoscaling, and aggressive caching.
I tried a couple of things (GHA, Dagger + Hertzner, Buildkite)
And Im just not too sure theres going to be any out of the box solution since my priority is essentially to minimize cost and maximize efficiency. Not really a great customer for any providers.
Im tempted to just get agent to build something out quickly with cloudflare workers + spot instances.
I also have some other nice to have requirements:
- ts/code over config
- locally runnable and testable
- preferably no lock in
- repeatable/reproducible
A buddy of mine runs his whole CICD setup off an old gaming desktop. They use tailscale to connect to their hosted infrastructure and set it up as a GitHub action runner.
For a solo dev this might be the way to go.
My wife uses my old gaming desktop for her ux design work as well.
And I was thinking of using the gpu to run some tts models.
Now to just figure out a way to run it all on windows and have it auto start when she logs in.
https://docs.drone.io/server/provider/github/
Very easy to stand up, does just fine. Definitely doesn't have the "library" of prebuilt actions that GHA does, but for the most part... I consider that a plus.
Otherwise it's very similar in concept - define actions in a yaml file, run commands on an image, webhook integration with most repo providers.
I run it on some old hardware locally (k3s cluster on old machines) and it outperforms the 1000 minutes from GHA easily, and costs basically nothing but some maintenance and time.
I've been keeping my eyes open for something new in this space since Harness bought it, though - so if other folks have recommendations I'd be interested in alternatives.
Best decision we ever made
[1] Ed Zitron speculates the actual prices with token based billing for heavy users will be something like 10x the subscription price, but this seems high.
Although I would also point out that OpenAI recently tripled the amount of Codex inference you get per month for £200 (and to head off the suggestion, this is distinct from their current 2x promotion on £100/month plans)
Neither of those is how much it actually costs the company selling the service. And I have feeling they are running at loss here so the play is "get everything possible using LLMs then jack up the pricing"
Inference is cheap but training is quite expensive. Plus all the money they've invested and keep investing on hardware, data centers, etc. And evidently they also need to make a profit at some point.
Maybe from the perspective of traditional, turn-based chat. But when you start having developers command an army of agents that work around the clock, those cheap tokens start adding up fast...
I think the margins have to be a lot higher than that in order to give investors the return they're expecting, to continue the never-ending training treadmill, and to build more and more datacenters to accommodate people basically DDOS'ing the GPUs in order to run their workloads.
Yes, in theory what you said makes sense. But the tightrope these companies have to walk is that the per-token costs still have to be low enough that developers and companies don't just say "ehhh I guess we can still do all this work the old-fashioned way" but ALSO high enough to cover the massive expenses AND astronomical returns everyone's expecting.
If prices go up, I suspect a bunch of folks will jump to cheaper, less capable models instead of eating the added cost. The whole value proposition of AI in enterprise is around cost-cutting, so that mentality is likely to persist when choosing which model to pay for.
There’s a difference between criticizing a company and just swapping names for insults. The former can be useful, the latter just turns the discussion into noise. If you’ve got a point about Copilot or the review feature, make it. Otherwise it’s hard to see what anyone is supposed to take away from “ShitHub” other than childish shit-posting.