Cognitive warmup. It was in January 2023 when Buzzfeed Inc. CEO Jonah Peretti announced in a memo to staff that the company was making a hard pivot to AI. Two points to remember. First, this happened two months after OpenAI’s ChatGPT caught everyone’s fancy. Second, many years before the word “slop” became a real thing, much to the annoyance of AI bros.
The transition Jonah Peretti was talking about—using AI to generate articles and other content on the website. Turns out, the readers didn’t want that. Buzzfeed has posted a $57.3 million loss in 2025, underlining a disastrous infatuation with Artificial Intelligence.
ALGORITHM
- Rollout of Meta Avocado may get delayed
- The visual challenge for AI chatbots
Meta’s Avocado needs more time to ripen
Perhaps the choice of the name for Meta Platforms Inc.’s next AI models should’ve been a giveaway—it is incredibly difficult to time when an Avocado is ripe and ready. It is now being reported that after spending billions on its AI pursuit through last year, Meta feels that Avocado models are still behind rivals including Google Gemini. The challenge here might be that previous testing pegged Meta AI on a par or better than Gemini 2.5, but Google then rolled out Gemini 3.0 late last year, which itself was a step forward.
After all, it was only last year that CEO Mark Zuckerberg and the Meta AI communications pitch had called the Llama models “most advanced in the industry”. For Avocado to succeed them as it is supposed to, it’ll have to be a big leap forward. I’m sure there will be iterations in different parameters, once the Avocado family is ready for primetime.
A chatbot’s visual context
Anthropic PBC’s Claude can now create custom charts, diagrams and other visual elements as part of a chat. The latest update automatically adds the visual element based on context. The rollout is in beta, for all subscription tiers, including the free tier (at least for now).
Soon after, OpenAI said ChatGPT will generate dynamic visuals when you ask it to explain select scientific and mathematical concepts, including the Pythagorean Theorem, Coulomb’s Law and lens equations. ChatGPT will respond with interactive visuals when asked about more than 70 concepts, with expectations that support for additional topics will be added in due course. The visuals feature is available to all ChatGPT users, regardless of subscription status.
THINKING
I have absolutely no idea why AI bros are so obsessed with innate biological necessities and human habits. Recently, OpenAI CEO Sam Altman went on and on about how much energy it takes to train an AI model and how much energy it takes to train a human. Of course, someone had to catch that cue and carry it forward. I’m not entirely surprised it is Perplexity’s Srinivas, who has been uncharacteristically less talkative on X (it is generally not a bad thing when AI execs post less on social media) these past few months.
But he’s back, with a pearl of wisdom that “A computer used to be a person” and that their new product called Perplexity Computer “never sleeps”. Because, apparently, “for a single person to build a 1B company, what’s the single biggest disadvantage they have? They have to sleep at night”.
For someone to start a monologue on X equating computers with humans and then set the context by saying “when astronomers had questions, their apprentices were the ones who actually did the work to try to find answers” and therefore they are the computers, is actually not making any sense at all. I’ll see if you can.
Srinivas is pitching what he calls a persistent digital proxy, which can turn a Mac into a locally run AI system.
- It works with a Mac Mini that runs continuously, merging a user’s local apps with Perplexity Computer. It runs on Perplexity’s secure servers.
- Every sensitive action requires a user’s approval, with a logged-in full audit trail, and a killswitch.
- A user can control this from anywhere, while the computer works from the desktop.
There’s snake oil. And then there’s this. “Everyone talks about ‘the first one-person, one-billion-dollar company’. It’s a meme,” Srinivas says. How many of you have finally stopped falling for these grandiose, philosophical and specious AI pitches?
Neural Dispatch is your weekly guide to the rapidly evolving landscape of AI. Want this newsletter delivered straight into your inbox? Subscribe here.
