Yajan

@paulg people who actually know what the best is tend to ask quietly and very specifically

It's a rough combination when people are simultaneously overreaching and uninformed. They want the best of everything, but they don't know what the best is, so their demands are simultaneously strident and random, like a set of vectors with large magnitudes and random directions.

@ChatGPTapp @grok explain the context

AI agents are wild gave one 3 hours of research work it finished in 20 minutes i spent the next 40 minutes verifying it didn't hallucinate so basically i have a very fast intern who confidently lies sometimes and i can't fire them

@yashhq_22 hypothetically, yes

Hot take: claude is still better than codex at coding if you just excuse the token limits.

entropy explains AI hallucinations better than "it's a training problem" ever will. the physics: entropy = systems drift toward maximum disorder unless something actively fights it every token a model generates pushes the probability distribution one step further from the signal 🟢 first 50 tokens: model is anchored to reality 🟢 first 200 tokens: mostly coherent 🔴 beyond that: you are watching entropy compound in real time hallucinations are not a data problem they are not an alignment problem they are the second law of thermodynamics researchers keep throwing RLHF and bigger datasets at it you cannot fine-tune your way out of a law of the universe

@tibo_maker so the actual strategy is just... be real and be good at what you do? wild how that keeps being the answer?

the researchers building LLMs don't fully understand how their own models work but sure, your 4-step GEO framework has it all figured out 🤣 everyone on the internet has a theory. do FAQ sections. get more citations. add reviews. make your content comprehensive - the list goes on but nobody can show you results actually attributed to any of it because attribution doesn't exist yet. and it won't until the people building these models can explain why they surface what they surface 🤷🏻 so what do you actually do with that? stop chasing GEO hacks ❌ your best bet is to do such a good job from day one that the internet builds a data bank about you real content pieces, real presence, real mentions - stuff LLMs will naturally pull from and make sure your SEO foundation is solid and that's why we and 2,500+ websites are so bullish on Outrank 🚀

@arena @OpenAI rubbish

GPT-5.5 by @OpenAI is now live in the Arena, landing across multiple leaderboards. Here’s how it ranks by modality: - Code Arena (agentic web dev): #9, a strong +50pt jump over GPT-5.4 - Document Arena (analysis & long-content reasoning): #6, on par with Sonnet 4.6 - Text Arena: #7, Math #3, Instruction Following: #8 - Expert Arena: #5 - Search Arena: #2 - Vision Arena: #5 Strong, well-rounded performance, especially in Code (+50 pts vs GPT-5.4). Congrats to @OpenAI on the release. Full category breakdowns by modality in the thread.

@hridoyreh happened to me twice

GoDaddy can take your domain:

@Kalshi tired of winning

BREAKING: S&P 500 hits new all-time high

last week in tech was based.

@deepseek_ai openai and anthropic charging 100x more for roughly the same benchmarks. at what point do we stop pretending closed source is worth the premium?

🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. 🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice. Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today! 📄 Tech Report: huggingface.co/deepseek-ai/De… 🤗 Open Weights: huggingface.co/collections/de… 1/n

@sama Is it agi?

GPT-5.5 is here! We hope it's useful to you. I personally like it.

@TheAmolAvasare ship, get ratio'd, revert.

For clarity, we're running a small test on ~2% of new prosumer signups. Existing Pro and Max subscribers aren't affected.

@thsottiaux the fact that twitter outrage is now the fastest way to get a feature back says everything about how product decisions get made in 2026

I don't know what they are doing over there, but Codex will continue to be available both in the FREE and PLUS ($20) plans. We have the compute and efficient models to support it. For important changes, we will engage with the community well ahead of making them. Transparency and trust are two principles we will not break, even if it means momentarily earning less. A reminder that you vote with your subscription for the values you want to see in this world.

anthropic yanked claude code from the pro plan and users lost it. so they brought it back. this is the real product feedback loop now. ship, get yelled at on twitter, revert. agile development in 2026.

@Kalshi @grok verify

BREAKING: Iran officially reopens Strait of Hormuz

@kenwheeler now show your usage

cunning

@claudeai somewhere in sf

Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.

@unusual_whales @grok is this true

NEWS: Anthropic is preparing for the release of its new Opus 4.7 model, per the Information