Thanks for building and sharing, looks cool and is very entertaining.
I had similar idea for people to code poker playing bots and enter tournaments versus each other, this was pre-llm, however.
It would be fun if you hosted a 'tournament' every month and had each of the latest releases from the major models participate and see who comes out on top.
Or perhaps do open it up to others to enter and participate versus each other - where they can choose the model they want to build with and also enter custom prompt instructions to mold the play as they wish.
There have been a handful of these threads on HN lately, I thought I'd chime in as someone on the other end hiring engineers, recently.
- If you're using the new "AI Tools" to auto-apply, especially if there are additional questions asked in the application that you use AI to auto-fill - it's relatively easy to spot, and is an immediate disqualifier for me. (fwiw - it may seem like it's providing good/unique answers, but if you get 100 applications from people using ~similar tools, guess what, many of the answers are ~similar or follow the same format)
Sure - AI is increasingly more a part of all our workflows these days, but I'm still hiring a human and so want to hear from the human.
- Speaking of applications with additional questions on them - if the application has those questions, answer them! The more thoughtful, the better. Why half-ass the first step of the interview process?
- There are hundreds of applicants for every job, plain and simple - a resume is not enough. Everyone has experience, and education, and skills, and it all just blends together. You need to stand out. Whether thats your GitHub contributions, a link to a personal website with writings or projects, personal side-projects or cool hobby hacks you've worked on - you need to have something you can point to and standout.
When I get several hundred applications for a role - I can quickly narrow down to a top 5% or fewer just by those that put in effort and had something to showcase.
I don't think you ever fully do. But, brute force.
Putting myself in an uncomfortable situation tends to make it slightly easier the next time, and then a bit better the next time.
I think, for me, the anxiety is the unknown of the situation and what to do or how to act and how people will respond. Getting some reps seems to help.
I feel similar in a lot of ways: stayed too long at dead-end jobs early in career, avoided going to startups (risk) or faang (scared?) for too long, failed to have a side-project of my own 'take-off' into something really substantial.
So I clearly have no silver-bullet advice, but I say 20 failures sounds like a good start :).
There are very likely nuggets you've learned within these roles and projects that you may not even realize, but will naturally come forth and be applied in a future project. Just keep moving.
I've also struggled with this, and I'm not sure there's a silver bullet answer.
But I think a lot of the common advice (too simple to work!) is actually helpful:
- Limit scope
As much as you think 'no one will use this without XYZ feature'; no one will definitely use it without it existing at all. Just make the basic functioning product and see below:
- Get real users using it early (release early)
Having real users (perhaps even paying users) using my project, relying on me for fixes/updates, seems to be a nice kick-in-the-pants to fix/work on the project. Even with minimal users or zero-revenue, having that feedback that someone finds what you're doing useful rather than coding into a void is really motivating.
- Build something for yourself
If you need/want to use it, you'll continue to work and improve the project to suit your needs. Theres some additional work likely to make it available for others to use, but if you're using the product yourself often - perhaps thats motivation to see how others could benefit from it as well.
Also realize - it's never 'finished'. If it's a project that you either (a) enjoy working on or (b) actually has users, there will always be more to do. So embrace that and find (a) or (b) early.
I had similar idea for people to code poker playing bots and enter tournaments versus each other, this was pre-llm, however.
It would be fun if you hosted a 'tournament' every month and had each of the latest releases from the major models participate and see who comes out on top.
Or perhaps do open it up to others to enter and participate versus each other - where they can choose the model they want to build with and also enter custom prompt instructions to mold the play as they wish.
If you walk this path, would love to chat more.