lain
@lain@fediffusion.art
pleroma
2.10.0
0
Followers
0
Following
Joined October 23, 2023
Posts
lain
@lain@fediffusion.art
fediffusion.art
> streaming site
> "write me a new interface that lets me download stuff, make no mistakes"
> 20 min later i have it
View on fediffusion.art
> "write me a new interface that lets me download stuff, make no mistakes"
> 20 min later i have it
0
1
0
Thread context
3 posts in path
Root
@lain@fediffusion.art
Open
@lain@fediffusion.art
Open ancestor post
Parent
@grillchen@brot.eus
Open
@grillchen@brot.eus
@lain can ai fill the holes in hour hearts? no ? didnt think so. trans-fatty acids can though
Current reply
lain
@lain@fediffusion.art
fediffusion.art
@grillchen *i robot robot voice* can you
View full thread on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
IRL ai assistant thing i did today.
1. got an email 'your subcsription to magazine x has been renewed'
2. forward the email to ai assistant
3. ai assistant messages me on telegram 'what should we do about this'
4. i say 'is all my data in the attached pdf? if so send them an email that i want to cancel.'
5. assistant sends the email and puts me in CC
6. magazine confirms the cancellation
View on fediffusion.art
1. got an email 'your subcsription to magazine x has been renewed'
2. forward the email to ai assistant
3. ai assistant messages me on telegram 'what should we do about this'
4. i say 'is all my data in the attached pdf? if so send them an email that i want to cancel.'
5. assistant sends the email and puts me in CC
6. magazine confirms the cancellation
0
1
0
lain
@lain@fediffusion.art
fediffusion.art
i love that the dream of open source software (you can just modify software to your liking, check if for security issues, add features, and so on) is now attainable by much more people, and will soon be available to anyone.
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
there's no need for any integrations anymore. want codex to use claude? just tell it to run claude in tmux and ask it to do the work.
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
just wanted to confirm that, while codex can by now write code faster and mostly better than me, it will never be able to do that for your code, that is, you, the reader, are 100% safe and you will not have to change how you work at all. your use case is so special that no amount of improvements in technology could every touch it. it needs the natural warmth of the human soul, and no clanker will ever take it from you. lucky you!
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
i'm currently maxing out 2 claude subs and 1 openai one.... claude are really too expensive. openai / codex is a much better deal
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
very minimal single user fediverse server written in zig with TDD in ~24 hours, by ChatGPT 5.2 xhigh.
https://github.com/lambadalambda/feddyspice
View on fediffusion.art
0
1
1
lain
@lain@fediffusion.art
fediffusion.art
Got the small x4, esp32 based ereader with a an open source firmware. Adding features to it with codex works pretty well, first thing: embedded image support.
View on fediffusion.art
0
0
1
lain
@lain@fediffusion.art
fediffusion.art
claude just told me that I have 3 one-week trials for claude code to give away. If you want one, tell me.
View on fediffusion.art
0
1
0
lain
@lain@fediffusion.art
fediffusion.art
it is suspicious to be living right at the time when we immanentize the eschaton
View on fediffusion.art
0
0
1
lain
@lain@fediffusion.art
fediffusion.art
> How do I feel, about all the code I wrote that was ingested by LLMs? I feel great to be part of that, because I see this as a continuation of what I tried to do all my life: democratizing code, systems, knowledge. LLMs are going to help us to write better software, faster, and will allow small teams to have a chance to compete with bigger companies. The same thing open source software did in the 90s.
https://antirez.com/news/158
View on fediffusion.art
0
0
1
lain
@lain@fediffusion.art
fediffusion.art
A video completely AI generated, with open source models, on a machine with 16GB VRAM.
https://www.reddit.com/r/StableDiffusion/comments/1qae922/ltx2_i2v_isnt_perfect_but_its_still_awesome_my/
View on fediffusion.art
0
0
1
lain
@lain@fediffusion.art
fediffusion.art
Very well made video about the somewhat mysterious phenomenon of 'grokking' in transformers, where they will quickly memorize all training data in a few hundred steps but still completely fail on testing data, then seemingly not improve at all for thousands of steps before suddenly generalizing perfectly: https://www.youtube.com/watch?v=D8GOeCFFby4
View on fediffusion.art
0
0
1
lain
@lain@fediffusion.art
fediffusion.art
new hobby: taking an open core project and asking chatgpt to implement the missing proprietary functionality
View on fediffusion.art
0
0
1
lain
@lain@fediffusion.art
fediffusion.art
using coding agents to decompile nintendo 64 games
https://blog.chrislewis.au/using-coding-agents-to-decompile-nintendo-64-games/
View on fediffusion.art
1
0
1
lain
@lain@fediffusion.art
fediffusion.art
https://simonwillison.net/2025/Dec/15/porting-justhtml/
A report of porting a library from python to javascript with codex, the chatgpt coding agent. It took ~5 hours and only needed one user intervention, when the prompt credits ran out.
I can confirm that GPT-5.2 has become MUCH better at autonomous work. I had it work on a hard problem that required it to constantly take screenshots and view the results, on an existing codebase, for over 8 hours without intervention.
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
https://github.com/DGoettlich/history-llms
cool idea, training LLMs only on data that's older than 100 years
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
the special fx people at corridor crew make a toy story short film with comfyui and wan. these are fully open source and run on local hardware. generative AI does enable completely new workflows.
https://www.youtube.com/watch?v=DSRrSO7QhXY
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
no idea why but non is really hard to train for. maybe there weren't enough gyarus in the original dataset?
View on fediffusion.art
0
0
0
lain
@lain@fediffusion.art
fediffusion.art
there's a huge amount of people who, upon learning that a bunch of bytes can learn a whole new style of art in 3 hours, based on a dozen examples, only feel despair and hatred.
View on fediffusion.art
0
0
0