Sabitlenmiş Tweet
idlevibecode.quest is "officially" live! With fun random events, you too can be a successful vibe coder!
English
Jamie (Taylor's Version)
8.9K posts
Jamie (Taylor's Version)
@TweetsByJT
Your friendly neighbourhood AI Dev Nerd. Podcaster, Gaming, Radio, Chemistry, F1 & Football (#WWFC). All views my own, most jokes stolen. He/Him
Loughborough, England Katılım Kasım 2008
444 Takip Edilen274 Takipçiler
Jamie (Taylor's Version) retweetledi
Uh-oh! Banned from arXiv because you you left some AI slop in your paper? It's okay, credit your LLM co-author and it'll probably be accepted into The Journal of AI Slop!
journalofaislop.com
Submit now!
English
Don't worry, as long as you credit your LLM co-author, we'll accept all your slop research at @JournalOfAISlop, the more hallucinations the better!
Thomas G. Dietterich@tdietterich
If generative AI tools generate inappropriate language, plagiarized content, biased content, errors, mistakes, incorrect references, or misleading content, and that output is included in scientific works, it is the responsibility of the author(s). 2/
English
@mynameis_rob @NaOHBartfield Oh yeah, now that's a blast from the past!
English
@TweetsByJT @NaOHBartfield The Newland Yield Reduction already exists, I hope I'm not inspiring undergrads to accidentally generate chlorine 😂
English
if you ever think youre dumb, just know that Twitter User NaOH Bartfield accidentally did a semi-wet gel transfer in lysis buffer instead of transfer buffer on May 7, 2026. foamiest Western ive ever done that's for sure.
keep your lysates
English
@mynameis_rob @NaOHBartfield Later cited in a series of papers (hugely popular with undergrads) as the "Newland Maneuver"
English
@NaOHBartfield I don't know what that is, but I once poured several litres of bleach waste into the acid waste instead of the bleach waste and the lab has to evacuated because of the chlorine.
English
i have made the mistake of talking about "the goblin problem" with my main @ChatGPTapp instance, with "memory" and "use previous conversations" enabled - now it REALLY mentions them in every conversation!
English
I've watched many podcasts/videos of opinionated people being softer and more considerate when there are kittens/puppies around (looking at you @theo).
I therefore propose that small fluffy animals be present at all high-stakes meetings/negotiations.
English
that settles it i'm an openAI stan for this generation of models
alex fazio@alxfazio
5.5 uses uv instead of pip by default on a clean environment
English
@initjean Is this just a harness problem? Compaction is essentially sending the current full context window with an instruction to compact - maybe there's a 5.5-specific improvement to the prompt that can make it better?
English
the GPT-5.5 honeymoon period is over for me
it’s broken for long conversations, and refuses to move away from bad ideas
while GPT-5.4 with compaction could go on forever without an issue
i should have watched the Theo video before. it honestly ruined my day
English
I'm naturally removed by distance from US stuff, but has anyone drawn the link between the "RFK-Trump Math" and the classic "Steiner Math" clips?
English
Played around with @ChatGPTapp's latest image gen model, I'm most impressed with the edit+tweak process. Many image models I've found overshoot on edits, not respecting the first generation. GPT-I-2 feels like it edits thoughtfully. Anyway, here's me as my WoW character.
English
#Eurovision fans!
I built a site where you can rank your favourites for each stage of the contest, join leagues with your friends or the community and compete to see who has the best Song Contest instincts!
fantasyvision.app
English
The document storage and graph aren't new by a long shot, but the way they're surfaced over time within the agentic loop is (from what I've researched). Early prototype (working with patches to #Openclaw) is on my gh: github.com/Popidge/memoria
English
Doing some sideproject work on #Agentic memory. Mapping the memory content as a graph, then every agentic step within a turn highlights nodes in that graph, similar to neurons in your brain. Above a certain threshold, the content they refer to is pulled into context
English
@Its_Nova1012 5.4 as your base version for sure. It's just a hella capable model. Then experiment with thinking effort for your token+speed/ability balance point. I tend to stick to High, but I do think Medium is good for a lot. 5.4-Mini will likely do well for pure boilerplate to save usage
English
@the_vibepreneur @OpenAI I don't think so, its more a band-aid. Plus you'd need some way to link a repo to a "chat app" thread/project. You can in the Codex web app, but not ChatGPT. That's the connection I'm really longing for!
English
@TweetsByJT @OpenAI Totally agree! The IDE ↔ LLM chat gap is real. What if there was a //context command that auto-injects repo context into the chat? Would that solve the immediate need?
English
One thing I'd love as a solo dev is a better integration between my ChatGPT projects where I discuss, research and ideate, and my repo I'm working on in Codex (or equivalent for Claude/Code). There's a disconnect in the two that could be a real help to solve @OpenAI (1/7)
English
Sadly, to get that level of integration, OpenAI would need some sort of API to interact with your ChatGPT account and it's chats. It's not something I can make (believe me, id give it a damn good go). So @OpenAI @OpenAIDevs - this is my feature request! (7/7)
English
Vice-versa argument for "just get openclaw to use codex as well", although that's closer to the desired experience, but requires setup especially to replicate the chat side. (6/7)
English