My former co-founder, Tom Ray, came up to London from Brighton for lunch with me in Hackney Wick today (I live in HW). He lived here around 10 years ago and the last time he was here was around 6 years ago. I gave him a little walking tour - he didn’t recognise the place, but loved the retained Berlin-like vibe. We had awesome catch up. Such a good human.
@petergyang I’m reaching here, but are you suggesting that ChatGPT, Grok, etc are effectively becoming your browsers? If you’re using them “on your computer and browser”, will we need Chrome, Safari and other browsers in the near term?
Last night's "Understand the Chinese Frontier AI Models" event hosted by Xiaowei Zhu and the @ShallWeTech team in the Amazon Web Services (AWS) London office was excellent.
Here's what I took away:
- Chinese AI model capabilities are ramping fast, and costs are currently very attractive relative to the US frontier labs
- Adoption is growing quickly, too (massive download volume in Germany) - and increasingly based on cost and capabilities; data residency, data security, sovereignty and related governance are increasingly solved problems
- Labs in China aren't altruistic - they aren't providing intelligence for nothing; they're monetising just like the US labs i.e. via direct web subs and API use in-country while shipping open weights for the rest of the world to build cred (and it's working)
- The adoption of Chinese models outside of China is on the rise - and safely via infra via Amazon Bedrock, Fireworks AI, OpenRouter and more
- Chinese models are generally cheap now, but expect pricing to increasingly swing in line with US frontier lab model costs, especially as Chinese model parameters grow and [potentially] cost more to serve
- If it feels like there's a glut of open weight models now, there are many, many more coming - from everywhere. China is playing this game well now, but we'll see a large number of fast followers imminently
Huge thanks for a superbly insightful get-together.
BTW, according to Zhu, there were 600+ registrations, 130 accepted and, I'm guessing, ~100 attendees. Not bad really. Free events typically see large numbers of no-shows. Those folks who signed up and didn't make it missed out. It absolutely fascinating.
Event details/link 👇
Note: I'm an early learner here, so I may have gotten a few things wrong. Feel free to course correct me.
Agreed. I’m a Substack newbie and totally turned off by the spamming. Did I ask for that? Must have been a hidden preselection. So, I’m baking an alternative platform citizen journalism. What would it take you to switch to an alternative? What would you want that Substack doesn’t have? Or not want?
@rampagingruss If there was a slick alternative to Substack, would you switch? I’m baking something. It’s not another social feed model. The foundation is good long form writing, with some special sauce.
Listening to @DeadKennedys in the gym. One of my go-to bands in the early/mid 80s.
Soup is Good Food just came on.
“We're sorry, but you're no longer needed
Or wanted or even cared about here
Machines can do a better job than you
And this is what you get for asking questions
The unions agree
Sacrifices must be made
Computers never go on strike
To save the working man you've got to put him out to pasture”
Could have been written today - only a few artists will go there now. Jello and the band were so on fire 🔥 🔥🔥
Lyrics from: @Musixmatch Thank you.
ExoCompute dev updates:
Cleaner launch teaser
Pre and post launch briefs require web research and 3+ verified sources before publishing
For fun, the countdown clock now updates by the second
And, a big one, added free Launch and paid Explore plans!
Give it a spin: https://t.co/YrMt1pFpWI
How I wish I had this with my last venture. Not necessarily me spinning up a bunch of agents, but my engineering team shipping product like this. Even better, as a non-tech founder, I would have been actively building, too (like I can now). What a great time to be alive and building.
Two very different founder conversations in the past week.
One’s nearly 10 years into his build in an incredibly crowded and competitive space, but with lots of room for growth (and new entrants coming thick and fast to reinvent with AI-native solutions). His position re: AI to craft new products, improve current workflows, serve customers? “We’ll start in the new year”.
The other founder is about 8 months into his build, which is an AI-native services business solving problems fast, zeroing in on outcomes-based compensation, and ramping fast.
I’ve done enterprise software sales and built my own software ventures, but I’m an old skool services guy at heart - or I started in services anyway. Specifically, I did digital media planning and buying, laced with frontier tech at the time (1999 to 2008).
Gotta tell ya, I’m excited by services again. Services with today’s technologies??? Wow!