randomcatuser

i mean, to be fair, these are professional researchers.

i'm very inclined to trust them on the various ways that models can subtly go wrong, in long-term scenarios

for example, consider using models to write email -- is it a misalignment problem if the model is just too good at writing marketing emails?? or too good at getting people to pay a spammy company?

another hot use case: biohacking. if a model is used to do really hardcore synthetic chemistry, one might not realize that it's potentially harmful until too late (ie, the human is splitting up a problem so that no guardrails are triggered)

what about crypto voting schemes? zero knowledge and all that

if we assume the user connection is secure (ie, about as secure as banking), can we have secure internet voting?

By publishing the poison fountain, you are making it so that researchers will have to invent techniques to "de-poison" data, perhaps contributing to long-term AI advances in intelligent data filtering while training

And secondly, why would you want worse LLMs? Seems less useful that way

This is so cool. A question on the service mesh - is building your own typically the best way to do things?

I'm new to networking..

A one-click avatar replacement (for when you detect you are in an AI interview) - that'll be interesting huh

The fun part is that then, the interviews can look way different than today -- e.g. the robot interviewer can demand proofs of the user's skill, etc.

It can even be confidential (ie, robot interviewer <> user's agent in a black-box room) so that they can share data

Imagine you have a function f(user_profile) -> decision

You can run f in a way that respects the user's privacy (and also hides the details of f from the user).

Companies get ~10x more data from each interviewee

Interviewees don't need to even show up

Sounds like a good deal to me!

I don't know if you're joking, but here are some answers:

"The mind is not a vessel to be filled, but a fire to be kindled." — Plutarch

"Education is not preparation for life; education is life itself." — John Dewey

"The important thing is not to stop questioning. Curiosity has its own reason for existing." — Albert Einstein

In order to think complex thoughts, you need to have building blocks. That's why we can think of relativity today, while nobody on Earth was able to in 1850.

May the future be even better than today!

i'm curious, how does it work? does it read the db directly or use applescript to get the notes?

Yeah, one way to think about it is like... protocols restrict things, so that people can expect the same stuff.

With a traditional API, people can build it any way they want, which means you (the client) need API docs.

With MCP, you literally restrict it to 2 things: get the list of tools, and call the tool (using the schema you got above). Thus the key insight is just about: let's add 1 more endpoint that lists the APIs you have, so that robots can find it.

Example time: - Build an MCP server (equivalent of "intro to flask 101"): https://developers.cloudflare.com/agents/guides/remote-mcp-s... - Now you can add it to Claude Desktop/Cursor and see what it does - That's as far as i got lol

Oh this is really cool! I was building out a bit of this with Restate this past week, but this seems really well put together :) will give it a try!

You could use Framer

https://landerx.framer.website/

you are a free man, just leave! imagine you only have 1 month left to live, how would you go about maximizing your life?

and do that

Wow, anyone have examples in SST? I just started learning CDK today, and this seems slightly better?

yeah the demo I saw was: https://x.com/livekit/status/1870194686532694417

But searching for "voice detection with pauses", it seems there's a lot of new contenders!

https://x.com/kwindla/status/1897711929617154148

this one is a fun approach too https://x.com/zan2434/status/1753660774541849020

Exactly what I did today! (for Codex!) The output here is actually slightly better!

I bet in the next few months we'll be getting dynamic, personalized documentation for every library!! Good times

The comments on Bluesky are really cool, did you implement it yourself?

I wonder what other things Bluesky has!

this is the most brilliant piece of writing i've ever seen -- reminds me of Godel, Escher, Bach.

Would love recommendations for other things like this where the style is one of discovery!! (like math textbooks that make you feel like you invented something?)

yes, and it's on a per-layer basis, I think!

So if the model has 16 transformer layers to go through on a forward pass, and each layer, it gets to pick between 16 different choices, that's like 16^16 possible expert combinations!

to be honest, OpenAI seems a bigger threat than Salesforce in the next 10 years

I have to say, the examples are really good: https://github.com/restatedev/examples

For someone (me) who is new to distributed systems, and what durable execution even is, I learned a lot (both from the blog post & the examples!). thanks!

ChatGPT also helped :)

neumorphism!

You know what you should do next? Out-innovate Vercel & offer even more cool features!

Some ideas:

1) Live object state! (like liveview/instantdb!)

import {useLiveState} from 'zercel'

const [state, setState] = useLiveState({whateveryouwant}, {name: '/roomId/keyId'}) // will sync automatically in realtime between clients!

2) Storage!

import { storage } from 'zercel'

storage.put("hello", file)

Inspirational, thanks for the open source code!

haha yeah... this is just for fun though!

one funny feature: extending other people's conversations

Oh by the way, this is almost a pure install of Miurla's open-source Morphic!

https://github.com/miurla/morphic

A great jumping-off point for building custom experiences, since it's got a solid implementation of AI SDK (by Vercel).

Claude gave me some write-ups of how AI & tool calling is implemented in the codebase (added to my fork)!

Wait, what's the difference between using GRPO and traditional fine-tuning of Qwen using your provided dataset?

Would be super interesting to see which one is more data-efficient!

Wait this sounds really cool, I want to start exploring Kubernetes! Can someone with more experience tell me why I should? (currently I just make all my apps on NextJS/Supabase...)

Is there a cost advantage vs using a more managed experience, say, Fly.io?

Supabase is open source, you can self-host it!

Haha yeah I recently learned this too.

this was AMAZING, i'd visit a site that was just a bunch of comedy scripts like this haha

Let's think about it.

If your website runs by uploading the files, that means the content must be determined by the files.

Use youtube/chatgpt: - "how do static websites work" - "how to code files on my computer"

Step 1. ALWAYS MAKE A BACKUP! (copy the folder somewhere safe). Or use git - you can learn the basics in 15 minutes.

Step 2. Edit the files. 1) First, you need a way to edit the files. Use Cursor. 2) If it's some text you want to change, use Ctrl+F to find the text & replace it with your thing. 3) If it's something else, you can ask the AI to change it for you.

Step 3. After changing your files, you can upload them to the hosting company again. You should see the changes!

Hope that is enough to get started!