1,353 karma · joined February 11, 2025
I like when programmers do creative, goofy stuff rather than spending all their time cranking out sterile soulless SaaS.
It’s what separates us from the machines. For now :)
Edit: to be clear for anyone reading, don’t do this, it’s a really bad idea. You need to patch your stuff, even behind a VPN. If you’re unsure, just google: “examples of security breaches of unpatched software behind VPNs”
I know developers love to build, but do you think:
1) self-hosting git provides any competitive edge to the business over letting someone manage it?
2) it provides so much value that you’re willing to fund engineers to build, secure, support this on an ongoing basis?
I’ve found the answer to those is No in both cases.
The same reason you wouldn’t build your own internal chat tool, you’d use Slack. And you wouldn’t bother self-hosting your own Jira or documentation.
Code hosting is code hosting, there’s no difference where it's hosted. There’s no slowdown in delivery with using GitHub - their March uptime was 99.5% which annoys some commenters but it’s fine. That’s 45 minutes downtime per month which is tolerable.
You would spend way more effort and money building a jenky self-hosted solution to end up with a worse result.
Thanks for the tip, I’ll give this a try and see how it goes.
I’ll admit the UI has changed a lot recently and I find it more intimidating than when I was using it a year ago, so I mostly use Ghostty now.
I know this wasn’t the point I was supposed to take from your comment but I’m liking this idea
That will tell me how confident you are.
$10k out of $500k would say you’re toying around.
$10k of $20k seems like you believe it works.
I have a feeling I know the answer, but wanted to check.
Because what it looks like is you asked Claude to create a stock picking newsletter based on signals that retail commonly thinks are bleeding edge but institutions have been using for two decades and probably don’t give much edge anymore (AI analysis of news sentiment…)
And you backtested from 2012-2025, one of the longest bull runs in history. Of course your strategy of “buy NVDA” will win in that window.
Which makes me wonder: have you ever worked at a hedge fund?
I have to say I’m tempted to subscribe and have Claude manage an imaginary portfolio based on your recommendations and publish that. Because to be frank, I’m fairly confident your recs won’t generate alpha.
It’s usually because executive management bakes hyper growth into the assumptions because they really want the biz to grow, then it becomes marching orders down the chain as it gets misinterpreted in a game of corporate telephone.
“We need to design this for 1b DAUs”
Then 1) that growth never happens and 2) you end up with a super complicated solution
Instead, someone needs to say, “Hey [boss], are you sure we need to build for 1b DAUs? Why don’t we build for 50m first, then make sure it’s extensible enough to keep improving with growth”
> But nobody was optimizing for cost. They were optimizing for their next promotion. Each rewrite was a new proposal, a new design doc, a new system to put on a resume. The incentive was never to pick the boring, correct choice — it was to pick the complex, impressive one.
...I guess it could be possible nobody thought about cost at all, and this was all misaligned incentives and resume-driven development, but I find that kind of hard to believe? As someone who has made cost mistakes in the cloud, this claim seems a bit silly.
Not to detract from his experience, but I didn't actually see much payments experience at all on his resume, so I'm curious why he's branding himself as a payments guru. Kind of tech content creation fluff, I guess.
I thought inference was cheap so there was little marginal cost of a new subscriber.
I imagine Anthropic is trying to see how many users they can push to higher tiers with these new squeezes.
I hate to say it but I imagine it will work.
It’s going to suck for me, because I had gotten used to ridiculously cheap tokens, but I guess the era of subsidized tokens is over.
Is there any marginal cost associated with a new subscriber?
I have always heard inference is cheap and the cost was in training, so I assumed any subscriber was making them money, just not enough to cover their insane fixed costs.
But I am just guessing.
Glad I’m not the only one!
I’ve been limited so often this week I’ve setup half a dozen token compression tools in my workflow and had to do a crash course in token optimization.
Of course, it seems to only slightly delay the inevitable and doesn’t really solve the problem.
Inside the case somewhere on mine there was a list of all the names of the people who worked on it. Was pretty cool.
Yes. When I was a young engineer, I was asked to design something for a scale we didn’t even get close to achieving. Eventual consistency this, event driven conflict resolution that… The service never even went live because by the time we designed it, everyone realized it was a waste of time.
I learned it makes no sense to waste time designing for zillions of users that might never come. It’s more important to have an architecture that can evolve as needs change rather than one that can see years into the future (that may never come).
I was looking at tree-sitter myself for this task.
;)
I just asked it to do a security analysis in a subagent of an unmaintained browser extension and then go fix vulnerabilities it found so I can use it without worrying.
What I’m reading so far seems to be:
-selective use of models based on task complexity
-encoding large repos into more digestible and relevant data structures to reduce constant reingesting
-ask Claude to limit output to X tokens (as output tokens are more expensive)
-reduce flailing by giving plenty of input context
-use Headroom and RTK
-disable unused MCP, move stuff from CLAUDE.md to skills
But I’d love to learn if anyone has any good tips, links, or tools as I’m getting rate limited twice a day now.
Fun read.
Some things will never change
Whoa.
-Stephen Jay Gould