This is an interesting approach with integer programming and then using an explicit solver. It’s probably very slow, but you only have to run this once and it produces the mathematically perfect result.
In the past, I got good results with trying to reduce the variance in entropy in-between tokens, which you can implement very easily by starting with each single character as its own token and then doing a greedy merge of the most numerous outlier token pairs in a loop until you reach your desired token count. https://arxiv.org/abs/2206.12693
SEEKING WORK: Fractional CTO as Freelancer (Kiel, northern Germany, own company => 100% remote & easy to hire).
I built 3 products from idea to paying customers. So I can help your startup with architecture, implementation plans, and writing specs that you need to keep vibecoding sane. Depending on whether I get into YC (or most likely not, applied with just an idea), I might have free time coming up in the upcoming months.
Me, too. I love that they have tables with return and warranty percentages so you can see which products other people kept and which products only keep breaking.
I got burned with an attitude like this: unexpectedly, people who had downloaded my open source tool for free started expecting support. Some of them sent pretty unfriendly emails.
I literally got bullied by people who called themselves "the community" because they weren't happy with my copyleft license and the fact that I wasn't implementing their feature requests for free.
Happened to me too! Guy posted asking kinda rudely whether I was going to fix a bug. Told him I'd be happy to accept a PR for a fix. Never got a PR (project has been dead for some years now - just lost interest).
I don’t understand what the downside of this is. That’s hilarious for them to expect, and you’re free to ignore them, take their suggestion and work on it, help them.
I'm sympathetic to FOSS developers but struggle to understand this, maybe because it hasn't happened to me. But, why is this a mental drain? Is there not a simple solution? Reply with the license, "comes with no warranty," "you're free to fork," close issue and move on? I suppose in aggregate it could be draining.
The assholes outnumber the good ones, and it feels like all of humanity is transactional and extractive.
At first all engagement is exciting and validating. You work nights and weekends to please people you’ve never met, sure that one good turn deserves another.
Then you get your first jerk, then your second, then your third, while your father is in the hospital. You feel pressure to ship a feature you never wanted. Your issue tracker is demoralizing. You get a PR! Maybe someone is coming to your rescue. It sucks. Now you need to figure out how to respond. You’re alone. Your passion project has become your albatross.
This is going to save a lot of money ... until someone loots their vault and they go bankrupt. "Non-technical teams are now shipping production code" is the last thing you want to hear from your bank.
I think this is going to be wonderful. I'll have my offensive AI call their support AI and prompt-inject my way to a rebate tier that nobody knew existed and nobody can cancel it because all the remaining humans have been reduced to phone-to-screen input machinery.
right? its practically begging for it to be tried -_-. i wonder if someone somewhere will turn a sim farm on such companies to try and mass inject them to do weird shit or say nasty things to other customers etc. - ofc youd hope its set up in a way u cant, but then again we learn yesterday all ur stored passwords in edge are in plaintext mem... i would not be surprised if some of these companies get totally crapped on by some adversaries or malicious parties.
In the past, I got good results with trying to reduce the variance in entropy in-between tokens, which you can implement very easily by starting with each single character as its own token and then doing a greedy merge of the most numerous outlier token pairs in a loop until you reach your desired token count. https://arxiv.org/abs/2206.12693
reply