Richard Foster-Fletcher

@SkaleNetwork @PGATOUR Congrats @SkaleNetwork team

The @PGATOUR just teed off onchain ⛳ Introducing PGA Tour Rise Free-to-play. Skill-based. Powered by SKALE. Game on.

Humans& just raised a $480M seed round at a $4.5B valuation. Thinking Machines had the team, the vision, and the valuation too. But now the team is departing. Could access to compute be the reason? It's GPUs that decide whether ventures can be built. Humans& needs its investor Nvidia to help them jump the queue. Is this the future of AI lab startups: you need capital, talent, vision, and a VIP pass to GPUs?

@Dhaunae This reminds of me other AI use cases like self-driving cars. One day people will realise that we are describing the build of a human, not a machine. And that the gap between LLMs and a human brain are gigantuous - comparing the two will almost be seen a humurous in the future.

That is a great observation. However, once models can update their internal representations on the fly—learning across sessions without explicit retraining—network effects will re-emerge inside the model itself, and collective use will compound rather than dissipate. The risk is that this scenario never materialises, or only briefly, if companies reach the point of no return first: by eliminating human know-how via indiscriminate layoffs, they lose the capacity to run core processes without AI. At that stage, dependence on AI providers becomes structural, and rollback is no longer possible—and the result is a disaster for both the economy and society.

AI has no network effects. When a colleague uses the same model, their queries do not improve yours. When a thousand employees prompt the same system, the system does not accumulate their collective intelligence. The organisation does not get smarter through aggregate use. Each session begins without memory of the last. Each user interacts with something that has already forgotten them. This is different from most platforms that became valuable at scale. Slack becomes more useful when more colleagues are on it. Google became more useful as more people searched, because aggregate behaviour informed relevance. Network effects meant that adoption by others increased value for you. AI does not work this way. Your prompts do not train the model your colleagues use. Their refinements do not benefit you. The institutional knowledge your team pours into these systems does not accumulate anywhere retrievable. It dissipates. The organisation invests cognitive effort into a system that treats every interaction as the first. The result is a strange inversion. Platforms that became dominant through network effects rewarded collective adoption. AI rewards individual skill. The person who learns to extract value from Claude is not contributing to a shared resource. They are building a private competence that does not transfer. Organisations are treating AI as infrastructure, something that improves with scale and shared use. It is closer to a tool, where skill matters and accumulation does not.

@SavannahFeder For a brief moment in history we thought AI commenting for you on social media was a good idea. Sorry, forgot we haven't got there yet. Also, LLMs sound like LLMs. Doesn't matter what anyone claims.

Today, we're releasing Claude Code for marketing. It does a marketer's work in minutes by browsing, clicking, and posting like a human would. The marketing hire is now optional:

If coding is now the wrong place to invest in your development, what's the right one? Learn to read systems. 1. Learn to see how the parts connect to the whole. 2. Learn to ask why something is built the way it is. 3. Learn to follow a decision to its consequences. Coding is commoditised by AI, architecting is more human than ever. Use the former, master the latter.

Congrats @Replit, this is really rather impressive...

MIT puts enterprise AI rollout at 5%. If you read that as AI failing to prove its value, please turn your attention to Replit. They just announced natural-language mobile app development. Describe the app, preview it on your phone, ship to the App Store. A year ago this took a team, weeks, and Apple process fluency. Now it takes minutes. What will you ship?

@paulnovosad @cblatts 💯

@cblatts It is a game changer, but it does annoyingly require totally re-imagining what your job is.

Approximately 1/3 of my X feed is people gushing about Claude code. I’m already an intensive ChatGPT user so I am open minded. And I will try it. I can't help but wonder: 1. Why do most of these posts sound like they were written by their AI? 2. Is this a viral marketing campaign? 3. Is this just the Twitter algorithm running wild? 4. Why don't I understand from these posts what these people are actually doing with Claude? Why is it all in vague gobbledygook? They talk about tasks in weird jargon, and it's like they're speaking a different language. I really don't understand what 5. Can someone explain in plain English what I would as an academic would do concretely with Claude code? We are already testing it out to clean and analyze basic survey data where it does okay. I'm going to be trying to play around with some new theoretical models, adapting IO models to criminal firms where ChatGPT has been doing ok. Will Claude code do better? Anything else I should be thinking about kind of work they're doing.

I've mentioned Claude code to several of my network. They always reply with the same question, "What should I do with it?" I've built 47 apps with Claude, all helpful, all useful. Everyday I have more ideas. Then there's posts like this saying, what do I do with Claude? Catch up...fast!

@PocaiJan Thank you. Please do check out my writing, Jan, if you have a chance: fosterfletcher.com/newsletter/

@RFosterFletcher Wow, that is interesting and paradigm shifting observation

@dhinchcliffe Curious, Dion, how many hours do you think you've spent using LLMs?

My take: I now believe AI will make knowledge work so cheap + plentiful that the only premium left will be “humanity”. Expect niche luxury markets for certain ‘white glove’ types of human work/experience: Talent sourcing, building communities, live person customer service, etc.

The Gemini growth story is interesting but the data measures something specific. Traffic to gemini.google.com grew from 5.7% to 21.5% over twelve months. Usage inside Workspace does not appear in these numbers. The growth comes from users seeking out AI directly, not from being funnelled through existing productivity tools. Copilot, despite being embedded in Windows, Edge, and Office, declined from 1.5% to 1.2% in the same period. Distribution does not automatically translate into engagement.