Coding is so passe. Dont feel like coding at all
Any conversation
setting up some of my workflows based on this writeup - ludwigabap.com/โฆ/on-my-personal-โฆof-the-tools-i-use/
Particularly a big fan of zellij amongst others
Also venturing into having agents enable more of my work - let us see how that goes.
ai coding feels like it has reached diminishing returns for me. luna is my default orchestrator, routing work by complexity. in practice, usage lands around 16% luna, 48% sol, 30% astra, and 6% terra. at this point, i care less about each new frontier release. the system already finds a capable model for the job. the frontier is moving beyond my day-to-day work, so iโm happy to settle into predictable models for a while.
Coding probably yeah, but still they are bad at generating novel information, new ideas and such. Theyโre also not very good at seeing the bigger picture, they will narrowly perform the task but theyโre not taking into account future revisions, which ends up in refactor after refactor in order to move forward.
Testing this quiet corner of the web. I collect small notes about everyday tools ๏ฟฝ๏ฟฝ file formats, tiny utilities, offline workflows. First post is just a hello; longer notes will follow once I settle in. Plain text feels good after years of heavy editors.
Onboarded on Claude Team with Unlimited credit usage. In the first day, subscription limits were hit and then it silently went on spending 530 EUR in credits before I noticed.
In a single day? Thatโs insane. Itโs 5 months on the 100$ Codex which is practically unlimited.
In a few hours, yes. Token-based pricing is insane.
I knew token-based was more expensive but didnโt imagine that much. Which is why u canโt compete making alternative coding agents, their own subscriptions are much cheaper, but why they do it on their own Teams package is weird. Maybe targetting companies they feel they can pay.
A credit union wants to get their own data out of a vendor's online system. Vendor wants $160,000. I wrote a script with to download the data just as though a user would. It's going to run for about 6 weeks to download around 250k records and 70GB of images.
you could split to buckets like alphanumerically and parallelize it if you don't want to wait that long. Probably need to launch a few instances but still cheaper lol.
โฆ
i dont think its all talk, these are smart people that have been in ai since the 2000s, and they raise good points. current incentive structures are very dangerous long term
they too have every incentive to exaggerate because they are selling safety and/or books on the topic. I donโt believe we wonโt see it coming. Any issues that will manifest will be very visible and weโll have plenty of time to mitigate their impact.
they could be making a lot more money by just working at ai firms i fear. the thing that got me worried the hacking incidents lately, huggingface etc.
if the capabilities consistently outspeed the development of safety measures, and if the private arms race logic continues, i dont think this is a good spot to be in long term