Article, for reference

In perhaps an interesting philosophical counterpoint to the analogy, the value in LLMs lies in their non-deterministic approach, in opposition to Herbert's understanding of both mentat and machine—a perfect deterministic calculator of the universe.

The teams, in particular managed failure in fulfilling two essential elements of the mentat requirements—they lacked the extensive and continued training, and they failed to provide appropriate context. "A mentat", we are told by Hayt "requires data".

This is, I think, especially relevant in a highly mapped level of specificity to Dune in context of the #Claude3 world simulation conversations on one side, and the @HarvardBiz article on teams failing to produce meaningfully improved results with GenAI brainstorming assistance.

I understand why the #Dune movies cut mentats—the movies are already verging on too esoteric for popular media, but it feels like they missed out on free mindspace in the zeitgeist from exploring the idea of rejecting thinking machines in favor of human intuition.

Everyone needs a “third space.” A place separate from home or work.

Also, education should be CHEAP. I should be able to educate myself without considering it an “investment” Education isn't just about earning potential, d*mn it. It's an investment in nature, in the pursuit of truth and beauty! Accessibility should be a no-brainer.

Skeleton-of-Thought: LLMs can do parallel decoding Interesting prompting strategy which firsts generate an answer skeleton and then performs parallel API calls to generate the content of each skeleton point. Reports quality improvements in addition to speed-up of up to 2.39x.

Both, please

"I work with LLMs every day. I can always tell when something is written by AI. They won't take your job. They're just not that good."

For two hundred billion dollars, you could fund 40 years of nuclear fusion research and reshape the entire planet. But I guess killing the bird and marking its grave with an X is a pretty cool, too.

Back when we couldn’t figure out what to do with computers, in 1977

The challenge with #buildinginpublic is that sometimes you get so busy with building you forget to do the in public part. Nights like last night remind you why you do both. Invigorated and inspired by all the amazing people I met! Thanks @Expert_Dojo @thepocketcrm @rickyschay

@ThePeterMick @the_feather_app will build friendships for people; and communities for the world. We seamlessly understand you, find your community, and curate your relationship. All you do is show up IRL, and build community. No profiles, no online engagement, no echo chamber. We’re what

@SamNewby_ The FOMO and novelty is dying out.. I think two things happened: - most dabblers realizing AI is still rough around the edges... straight out of the box. It's not that good without tweaking, embeddings or using complex prompts. - ones that did figure out the technical part are

I like LangChain, and use it a lot, bit these critiques are all spot-on. Documentation and debugging in particular is maddening.

Sales Operations

The upcoming death of the Transformer. Stop. Hear me out: Million tokens in the prompt. Intrested? Meet the (might be) transformer killer: Hyena. More powerful than a transformer. [*] Order of magnitude cheaper than a transformer. [**] But is it scalable like a transformer?

My backyard game, The Plunge, failed. Here's the story of why I'm closing the business & what I learned: I was interviewing one of my clients for my ghostwriting agency this week & he said something fascinating. I asked him "What are the best decisions you've made in your

Look, I'm just a first time founder, so what do I know, but it seems like pissing off your most loyal users is not great strategy. Anyway, if you are looking to step out of the collapsing cesspool that is the social internet in the next few months, check out @the_feather_app