It doesn’t have to be $300/day worth of output tokens. It could be like $290/day worth of input tokens to teach both you and the model about the problem you are solving and then $10/day worth of output tokens.
I have been hungry to do more work from my cell phone. It's ridiculous to be forced to sit in front of a computer to work with AI.
My current solution is to have claude (--dangerously-skip-permissions) listen for messages in my slack DMs to myself and take action in response to those messages.
Once you've built a system that works well in production (and scales elastically, too, it seems), it is really difficult to switch out the underlying infrastructure. Makes sense about walking it back.
The problem with running your own servers in data centers as a startup is that elasticity is genuinely a difficult problem to solve if you don't have a large budget for unused compute, storage, and so on. As we are seeing in Fly.io's case.
Ultimately, my bet is that both startups end up as Heroku-like acquisitions for some large cloud company or another. I think that render will sell for a lot more because the value it provides is agnostic to the underlying cloud infrastructure.
I'm curious how significant a risk products like AWS Lightsail are to your business - it seems you are competing in the same market, but:
1. They have vastly different ongoing capital and cashflow requirements than you do.
2. They have all the leverage when it comes to the question of your continued operations on their cloud.
I'm also curious if they have already offered to just buy you out since you're clearly succeeding where they seem to just be treading water. (But not expecting you to answer this question. :) )
I read your review, and had a question so I thought I'd follow up here. You mentioned render.com as a competitor - does render host its own infrastructure or do they act as a go-between between their users and AWS/GCP/whatever?
My current product started off as precisely that kind of search engine.
User adoption is a huge problem - almost no users made it their default search engine because even programmers need to do non-programming searches and it's too easy to go into your browser, hit ALT+d, bang out your search query, and hit enter.
And because google and ddg do a good job on most programming related searches, they get to be the default search engines.
Because we were not using any numerical properties of the hash. We were not adding it to other hashes, seeing if it was greater than or less than other hashes, etc.
Literally the only thing we were doing was passing it between shell commands, helm charts, Kubernetes deployments and then back (if we needed to debug).
It sounds like you have a more attractive alternative in this case than to treat hashes as strings. Would love to hear it.
Hashes are NOT numbers in base 10 scientific notation, which is how the hash that I showed you would be interpreted by YAML.
The point is that this behavior is sporadic. It doesn't apply consistently across all git hashes, which is the real problem. It is easy to be caught unawares by this behavior.
Have had a similar issue when adding git revisions to YAML documents.
The problem is that if a YAML parser sees a string like this:
"0123e04"
It interprets it as a number: 123 * 10^4
Our hacky solution was to prefix the revision hashes like sha-0123e04, but still this was quite annoying.
After that experience, I have stopped using YAML for any of my own configuration. Have started preferring putting my configurations in code. And when I don't want that, have found JSON good enough for my purposes.
Have gone through startup school, and it is not at all how it seems to you.
We met companies that were making hardware and companies that had been building complicated enterprise products for years. I don't think we met a single MVP in a weekend type of company.
All in all, a valuable experience - at the very least it gives you access to a bunch of nice deals (AWS, GCP credits and the like).
Having the measurement allows us to relate it to other measurements that others have made. We may not be able to draw conclusions from the Stack Overflow measurement, but we can use it to contextualize information that we are presented with in the future.
Why do you feel as if it does more harm than good? (Genuinely curious.)
It's a measurement that I am certainly curious about.
I do not belong to that most popular demographic, and have a passive interest in that demographic distribution becoming more proportional to its global distribution.