infrecursion
3.8K posts
@HedgeyeComm If they fired Sam Altman and took him off the board, they would actually have a chance. No bullshit.
English
Hard pivot to become Anthropic
First Squawk@FirstSquawk
OPENAI TO MERGE CHATGPT, CODEX APP & BROWSER INTO DESKTOP SUPERAPP TO STREAMLINE RESOURCES & USER EXPERIENCE - WSJ
English
@HedgeyeComm Is this somekind of a joke or just average hallucination of a claude sheep? Anthropic does not even have a model that does audio or image. It's pretty dogshit experience almost anywhere.
English
@SynBio1 Yes, but not because of what you think is the cause. It will be because humans like you become completely irrelevant.
English
This is the core problem of the next 5 years in the AI scientist era: infinite hypotheses about which nobody cares
Sayan@thesayannayak
At this rate everyone’s gonna have their own app and zero users.
English
@hypertectonic @btibor91 It could be a completely different model.
English
ChatGPT/1.2026.076 (Android) adds an announcement that "Video in ChatGPT is here" - "Transform text and image into video with dialogue, soundtrack, and style."
English
@nayshins Start by deleting your own codebase, every line. That will get rid of so much slop.
English
@petergostev The post below is a valid criticism and why I can't take this benchmark seriously. It's just choosing my preferred model with extra steps.
x.com/FakePsyho/stat…
Psyho@FakePsyho
I have eyeballed some of the outputs and imho the results are not as clear-cut as being presented (and that's an understatement). For a lot of outputs, it's very debatable what's the proper way of categorizing them. It seems to me that each model has it's own default style when answering a bullshit question and this style is surprisingly consistent between different questions. What you're mainly measuring is the preference that each judge has for each particular writing style. My other complain is that there's no diversity in questions: you're just embedding a random unrelated or made up terminology in a somewhat valid short question.
English
BullshitBench update: The new GPT-5.4 mini and nano models score quite low. This screenshot shows OpenAI models only, on the full list would put GPT-5.4-mini around 40th place and Nano is around 70th place. Again thinking didn't help much at all.
Peter Gostev@petergostev
BullshitBench v2 is out! It is one of the few benchmarks where models are generally not getting better (except Claude) and where reasoning isn't helping. What's new: 100 new questions, by domain (coding (40 Q's), medical (15), legal (15), finance (15), physics(15)), 70+ model variants tested. BullshitBench is already at 380 starts on GitHub - all questions, scripts, responses and judgements are there so check it out. TL;DR: - Results replicated - @AnthropicAI latest models are scoring exceptionally well - @Alibaba_Qwen is another very strong performer - OpenAI and Google models are not doing well and are not improving - Domains do not show much difference - rates of BS detection are about the same across all domains - Reasoning, if anything, has negative effect - Newer models don't do that much better than older ones (except Anthropic) Links: - Data explorer: petergpt.github.io/bullshit-bench… - GitHub: github.com/petergpt/bulls… Highly recommend the data explorer where you can study the data and the questions & sample answers.
English
@gabrielchua Please put GPT-5.4 mini on ChatGPT. I want to use a reasoning model for many small to medium capability tasks like web search that doesn't require full GPT-5.4. I don't want to waste my limits using the full model. 5.4 mini should be available for selection, not as a fallback.
English
Now with `gpt-5.4-mini` and `nano` out, I put together a simple cheat sheet of the latest OpenAI models by use case.
Noticed at a few recent hackathons & meetups: some folks still default to `gpt-4o-mini` for LLMs and `whisper-1` for transcription.
Newer options tend to fit better now with much better performance. If you’re running into issues switching, lmk!
English
@ryanwinchester Yes, there are still people offering rides and riding on horse carriages. What's your point?
English
@krzyzanowskim No, sorry the worst codebase to work with are those by you, no exceptions. You're a shit coder.
English
the worst. literally the worst. no exceptions. the worst codebase to work with coding agents is the one generated by the agents itself.
usually, the best codebase to work with agents is the one that predates coding agents.
English
@MyNameIsOnni @LeahLundqvist It's likely satire. They're quoting from the marketing post. Dall-E 2 is better.
English
"Even with the same prompt, DALL-E 3 significantly improves upon DALL-E 2"
English
@jxnlco Why isn't mini offered as an option in ChatGPT? I want to be able to use a mini reasoning model for medium complexity tasks and search without having to spend 5.4 limits.
English
@alekzan @dejavucoder Opus is not even close to GPT pro. GPT pro is a different level altogether.
English
@dejavucoder I don’t think so. Opus could be GPT pro and sonnet vanilla GPT.
Mini and specially nano have never been good models.
They’re always behind of even identical open weight models.
English
@LokiJulianus "Yeah imma take everything happening out there and try to fit it inside my worldview, because as everyone knows I am the center of the universe" - You.
English
@lefthanddraft It seems humans like you will get replaced before you keep repeating the same dumb questions with no undestanding of how these things work.
English
This thing is going to find a cure for cancer before it stops falling for dumb tricks.
English
Consumer AI will be left to Google and Apple
Funny enough, both using core Google model technology and (likely for now) infra
I wonder if this changes the OpenAI hardware strategy
The Wall Street Journal@WSJ
Exclusive: OpenAI’s top executives are finalizing plans for a major strategy shift to refocus the company around coding and business users on.wsj.com/3N6CFyr
English
Spent all day vibe coding, fixing bugs in AI Studio, and polishing the experience :)
So much fun!
English
@Sauers_ Dude what tf do you have in your system prompts that you get these weird results all the time?
English
@ToFollowBrights @Zai_org Lmao how insane is your entitlement? This lab has single handedly released one after another SOTA quality open source models and the moment they try to raise some money (from their products not vc), they become OpenAI wannabe. Go fuck yourself.
English
Introducing GLM-5-Turbo: A high-speed variant of GLM-5, excellent in agent-driven environments such as OpenClaw.
Coding Plan Max: z.ai/subscribe
OpenRouter: openrouter.ai/z-ai/glm-5-tur…
API: docs.z.ai/guides/llm/glm…
English
@nlarusstone Lmao, so you're basically so far up in your arse that you think the dog getting better was some kind of scam someone pulled off to confirm your worldview?
English
The funny thing is AI already is accelerating drug discovery AND consumer biotech is going to be a big thing.
This just ain’t it
English