#vibecode

19 posts · Last used 11d

Replying to
@antimundo@mastodon.gamedev.place There's also this: https://github.com/pixelfed/pixelfed/commit/ecb0131d33e0f008a66b88177dcf80cb93aaef01 That's for something called "Kiro" so you can "Move beyond AI coding to agentic engineering". Laravel is apparently "An Agent Ready Framework". It seems this dansup person behind Pixelfed has gone full "AI", and it's not just commits with documentation for "AI agents" that make it very clear. If you just look at the sheer volume of commits from him the last few months you can tell that he's either taking large amounts of amphetamine & coding like crazy day and night or using some kind of automated system to generate code. #pixelfed #ai #vibecode
4
1
2
0
Replying to
@kolya@social.cologne Ive been running multiple VPS'es running multiple Claude sessions. Ive explained my current workflow a while back but #Mastodon search sucks. I cant even find my own posts I did a couple months ago. Anyway, my CLAUDE.md has directives to; a) I have a folder called DOCs any time I have a problem/challenge that is resolved (I call them thesis) #Claude writes DATE_Thesis.md b) I also have ai_index.md which starts with a prompt to regenedate the ai_index. The index has a filename and a brief summary of existing thesis. The purpose of both of these files is for the model to come up to speed on the project very quickly, without grepping shit all the time. c) I also have file called NEW where all the ideas / bugfixes I dont have time to fix go. So in promt I just say NEW: Fix global hunger and it gets added. But specifically, the idea I have, clanky though it is, is to use one of my #Wireguard tunnels as a master /DOC repository and all the instances running Claude can write to the library of Alexandria. I used to do the copy/pasta thing but it drove me up the wall. Though I still do it for x-server sessions. But it irks me and I may need to fix it soon. #Vibecode #VPS #Selfhosting
1
0
0
0
Replying to
@maxleibman@beige.party I'm an ex-Alpha geek computer pioneer with my brain turning to mush. I have a sign by my egress door "Put on pants" The gif below is some screens stitched together from my Ai harness engine. It took me 2 years to put together a PHP content manager (like WP, but way shittier)... ...this harness would take me something like 4 years to put together. I made it in 2 months (of interactive, non contiguous progress) this is V4. Oh its also in Python, a language I have 0% familiarity with. You DEFINITELY get the x10 You ask, where is the proof? The proof is in the recruitment ads, all the geeks are bitching how few jobs there are. #x10 #vibecode
0
0
0
0
This is why #Ai #Vibecode proponents are asserting you get x10 gains... ... At the beginning of the year, I have created a framework to display multipe servers I am monitoring... ... I can not very well display my infrastcrucute to the public... but just with a few prompt commands, the #LLM created an entirely synthetic map of a large network (If you are a #SysAdmin you will see its synthetic) but its good enough to show capabilities. Note, everything is spinny, animated and shit. #Monitoring #infrastructure
0
4
0
0
Wow, this is the part that many #AiAntagonists do not understand. Not because they are stupid, but because #Ai is not a tech they use. The models get CONSTANT upgrades. Every week or two, there is a new build... Ive just discovered that #Claude Code can spawn agents to do work. Specifically I often bounce between models Sonnet/Opus. One for coding, one of depth of reasoning. Ive designed a thesis and usually I just step down into sonnet and get that to #Vibecode it for for me... ... but today, the model did that on its own; "now spawn a sonnet subagent to implement rev 2 Spawning it now. The spec is self-contained, so the prompt's job is guardrails and definition-of-done rather than restating the design."
0
2
1
0
AI slop
0
0
0
0
Replying to
@jwildeboer@social.wildeboer.net Tell me you haven't reviewed #Vibecode without telling me you've not reviewed AiGen code. You don't review code, you review function. And if you're building mission critical code with Ai... ...then you probably also BASE jump and freeclimb.
1
1
0
0

Truns out, the Stochiastic parrot LOATHES vague languate...

...

  1. BANNED VOCABULARY "reasonable", "appropriate", "clean", "robust", "as needed", "production-ready", "comprehensive" — in code, comments, and self-assessments.

Needing one of these words means an unmade decision: make it or mark BLOCKED.

BOOM APES! GET YOUR SHIT TOGETHER YOU WEAK BIOLOGICALS !

#Vibecode

0
0
0
0
Vibe coders go vibe killers, because hey, what could go wrong if we just let morons with exacto knives start killing things to see what happens? Look, I understand it’s a lobster. Lobsters are not puppies with adorable eyes, I get it. There are still ethical standards which MUST be followed. Anything less is unacceptable. There is a boundary between science and sadism, and that line is called Ethics. Thinking that you can #vibecode your way into letting #AI take over a sentient being is the epitome of how ridiculous this simulation has gotten, and how terrifyingly high on their own supply the AI community, funded by #Altman, is becoming. These are not explorers on the frontier of new science, these are children who think living beings are no more important than a lego brick. https://apple.news/AxemAJ_LDTHObJFLbCUuR5g?highlight=Biohackers%20Attempted%20Neurosurgery%20to%20Control%20a%20Lobster%E2%80%99s%20Nervous%20System%20and%20Give%20the%20Controls%20to%20OpenClaw,%20and%20How%20It%20Ended%20Will%20Tell%20You%20a%20Lot%20About%20the%20Ethics%20and%20Competence%20of%20AI%20Bros%20These%20Days
0
0
0
0
https://infosec.exchange/@n_dimension/116851756334409459 n_dimension - @bsky.brid.gy That's a good question. "How do I keep the prompt consistent across systems?" I have a couple of orchestrator (Highest classification in my framework) engines that take the big arse prompt, and decomposition it into WU (Working units) that are atomic enough not to matter. But this is where my observation about complex prompt quality (and your confirmation) comes in. https://media.infosec.exchange/infosec.exchange/media_attachments/files/116/816/479/786/383/989/original/d4bd80a6900e0911.png #ai #promptengineering #vibecode
0
0
0
0
So here is an example how #Vibecode does not rot your #programmer brain but actually helps you come up with innovative solutions to new (#AI) problems; ISSUE: Local LLM models are frankly... Dumb. THESIS: Design a Multi Agent Consensus pipeline. That is, a proposed deterministic code/data structure (bash,python, JSON, etc) is run through a chain of Agents, each one might be wrong separately, but together, they are wrong together !!! 😀 Then the sausage at the end, is verified with the WWW. The loop repeats until something usable pops out at the arse end. Parameters: Set temperature to 0 in OLAMA (So models stick to the plan and dont hallucinate). Also, use Syntax linters and hard guardrails on destructive shit. Big plus: You don't burn compute at teh evil evil evil #Datacenters , you just crank your VPS CPU which as we all know, is not located at a #Datacenter but on a free range electric field emitting unicorn dust not pollution. #ImSoSmart
0
0
0
0
Replying to
@soller@floss.social Just the other day I was barked at and blocked, in the context of the #vim "controversy" For saying that if you insist on #Ai #vibecode free purity... ...you will be using a clay tablet for your computing needs soon. And here we are. I'm looking forward to a HomoSapiensLinux fork where every line of code is chiseled from a pure human bone with tools forged in artisan smithy... ...that's going to collapse in controversy because major contributors will be sprung vibecoding 😄 BTW: Any of you read that .md file before conniptions? Would be good idea to get some human #codemonkeys to comply with the directives there 😁
2
4
0
0
You've seen all posts