This company is in tech and doesn't have an old neckbeard devops engineer keeping the servers running. They just hire hot racists from SF. I know that code is dogshit. How do I short them? https://nypost.com/2026/09/25/tech/inside-the-world-of-california-startups-corgi-girls-unhinged-dating-demands-bizarre-posts-and-outrageous-online-behavior/

1 reply

0 recast

3 reactions

Opus 5.5

0 reply

0 recast

1 reaction

Some wise words from a very familiar technowatermelon: https://x.com/varunsrin/status/2102597409843142950

1 reply

0 recast

7 reactions

When you're building software that needs to operate flawlessly with the limited flight pass opportunities of the International Space Station, you need libraries you can... Boost is one of the key dependencies we’re proud to build on. https://github.com/boostorg

0 reply

2 recasts

14 reactions

this is it. this is the first production-level step to true AI. we're going to actually be able to tune neuron weights in realtime with this when applied in layers. not to mention the myriad of regular applications this will completely revolutionize. exciting times. https://typesafe.ai/blog/introducing-system-one-models-and-jev

0 reply

0 recast

3 reactions

Great Jev breakdown - it’s a decision model - output is schema (easy for code to understand) - crazy cheap + fast

0 reply

0 recast

2 reactions

maybe not so refined after all

0 reply

0 recast

1 reaction

0 reply

0 recast

0 reaction

My weird fever dream of the day. PurpleSky, an atproto PDS that uses Farcaster for auth, data storage, and search. Basically a sharded PDS, but using Farcaster. https://psky.org

4 replies

1 recast

11 reactions

i'm wheezing man. 😂 😂 😂 https://x.com/0xdoug/status/2099647215773901064?s=20

2 replies

2 recasts

6 reactions

we need a coding agent trained on mvps only

0 reply

0 recast

1 reaction

Welp it's almost 5:30AM, but I'm done! Visor has gone from 1 -> 3 worker nodes. Nodes service the processing, categorization and taxonomy of OSINT news. I'm still working through the metrics but I have *ALMOST* achieved a 3X in feed processing ingestion in the queue, which is pretty fucking insane given the overhead in orchestrating workers... and shipping I/O between the workers' remote locations across the globe. I'm so fucking nerdy that I might spend the next week just talking about load balancing algos I architected for this. 🤓🙈🙉🙊 Absolutely chuffed with it. Ingestion rounds used to take on average over 3 minutes (~188s), and we're now down to ~64s median. The TL;DR is that you get more news, faster. Now that this is done, I will be looking at onboarding many more news sources, as feed processing and ingestion was a massive blocker--now unlocked! It also means, I think, we're basically ready to open registrations to the public! HUZZAH! 😻🗞️🗞️🗞️

2 replies

5 recasts

10 reactions

me with Astra

0 reply

2 recasts

6 reactions

deepseek 4.1 flash is pretty good guys. i just might cancel my extra codex subs

0 reply

4 recasts

11 reactions

the hubris of the software engineer has been completely decimated by inorganic neural networks

0 reply

0 recast

1 reaction