@MatijaSosic if only wasp ran on val town
Notes
- @pspeter3 I played around with client side hono a bit, but it had more downsides than client side react and basically no upsides. I wasn't able to find a gradual of great path from server side hono jsx to client side
- I'm in SF Feb 26 - March 3. Who's in town?
- @pspeter3 I use hono jsx. I'm not sure if that's a hono component
- @jake_researcher I'd love to see some minimal example code!
- @Dobrenkz I may have to hit you back, realizing I'm traveling this week. Tomorrow is unlikely too sadly
- Nevermind Townie seems to have fixed it! https://www.val.town/x/stevekrouse/reactRouter7Example Full-stack react router! Pretty neat!
- I have two stacks I love for val town. Stack 1: hono hono jsx 100% server-rendered 0% client side js twindcss pages/ directory components/ directory lots of <form> Downside: lack of client js Stack 2: hono react jsx 1% server-rendered bootstrap 99% client side react twindcss…
Text may be partial · Read original
- @Dobrenkz Woo! Let's do it! When are you around for a first 90 min livestream?
- https://www.val.town/x/stevekrouse/microgpt
- @andrewchen i'm hacking on this for a company's finances
- i'm trying to understand @karpathy's microgpt here's how far i got in 2 hours - would love feedback on how i'm going about it! https://x.com/stevekrouse/status/2025323555471159458/video/1
- @EmreCoklar @deadheat_ Perfect, let's do it! When's good for the first 90 min livestream?
- Is a mixture of experts architecture and a model router basically the same idea? Feels like there's something interesting about where you make that 'routing' decision
- @glcst density implies a denominator database per ??
- @samwhoo Seems feasible in the same way that more transistors on a chip seems feasible right up to the nanometer, even without knowing too much about all the magic tsmc does If they can do it with a tiny model, why not with a bigger one? We also keep making smaller models smarter. I've…
Text may be partial · Read original
- @samwhoo Yeah I regret "likely" in my original post. I think "feasible" is what I meant or "within striking distance"
- @samwhoo 2 years seems too fast? You don't think this new chip company will be able to get this working for bigger models? We won't be able to make small models super smart? Seems like custom chips (in retrospect) may be obviously excellent idea to get order of magnitude or two…
Text may be partial · Read original
- RT @dbreunig: The state of coding agents can be summed up by this anecdote: Claude spent $20k on an agent swarm implementing (kinda) a C-c…
Reposted by Steve
Text may be partial · Read original
- 20k tok/s seems likely within 2 years (50 tok/s is most frontier models, today) we are in the dial-up era of llms both in speed and price remember when you used to buy minutes and texts on your phone plan? that's how this era of ai will look in just 2, let alone 20 yrs https://x.com/swyx/status/2025041392196878729
- Do I know anyone brave enough that I can teach coding to on a livestream? I want to explore what it means to learn to code in 2026 with AI What do you actually need to learn vs what can the AI do for you? How long does it actually take?
- @geoffreylitt @blaine Agreed! Thanks for asking the question and thanks for writing that up! Super helpful!
- @GustoHQ @tomerlondon replied!
- @tomerlondon @GustoHQ ah, changing my email didn't work (i confirmed the password field is correct) https://x.com/stevekrouse/status/2024887913901674981/photo/1
- @tomerlondon @GustoHQ Ok I set them to the same email address, but they still seems like separate accounts How do I switch between employee and employer?
- @tomerlondon @GustoHQ Ok will give it a go!
- @tomerlondon @GustoHQ Thanks! Do you recommend I change my personal email (employee login) or my work email (employer login)?
- Texting with my HUMAN DOCTOR on the One Medical app. She just hit me with a "You're absolutely right" and I'm shook
- @tomerlondon @GustoHQ wow! really appreciate the quick reply and care 😍 (shot a note to @gustohq in dms)
- Do I know anyone at @GustoHQ? Big fan of the product, but logging-in is such a chore. I'd love to give feedback on how to improve that.
- RT @geoffreylitt: 🎯 Agent orchestration is a data visualization problem
Reposted by Steve
- having fun with sonnet 4.6 this morning to vibe or not to vibe, that is the question https://x.com/stevekrouse/status/2024476273653711184/video/1
- @nfarina it's a joke because there are so many now
- would someone just make an ai gateway already?
- Ok I'm late to the party, but RLM (recursive language models) is very cool! Has anyone I know played with it? Any tips for getting started in JS?
- @eudaemonea ha! I wonder if that's a good solution I need this programmatically, so I'd need to host Claude code somewhere
- @cindyyhao_ @ryanguill @ExtendHQ yeah, or the equivalent to pass a tool to the claude sdk or openai sdk
- @EmreCoklar @intellectronica Thanks! I'd love a guide for the vercel AI SDK in particular. In part, that gives me confidence that the vendor understands me and my problems
- @cindyyhao_ @ryanguill @ExtendHQ thanks! I'm looking for a guide that looks like this https://x.com/i/status/2024192763424637063
- @mattgillard i wonder if there's something in js-land for this
