dcsan

Wish this project had more traction. Typechat with type checking could generate lots of synthetic data for model training too

The addition of "feet first" made this much more evocative!

For 100s of top level topics you'd need many curators. And this way they're generated on demand

the enshittification starts here. - Discord IPO plans forcing - new CEO installed by VCs - instantly adds spammy ad popups - closes down app partners

platform risk is real. twitter, reddit, twitter again ... if you build on top of another platform you're just collateral damage.

Time for shapes to get their own app and see if there is enough stickiness or if they were just a 'feature' all along.

the main benefit of hono seems to be you can deploy to cloudflare either as a pages app or workers. so this is kind of like a 'sveltekit for react'.

in which case it would be nice to see it extended slightly: tailwind client side and eg drizzle ORM - basically the minimum you need to really ship a client/server app

BBC tv or radio? Seems odd now to think back to a fixed time radio broadcast but I loved the radio show

I so fondly remember StumbleUpin, but I’m trying to recall what was so amazing about it. Was it just something of a novelty at the time or the autocuration of the decentralized web would still be relevant?

it seems like a few social media sites took over from the random delight of finding someone’s little weblog or side project.

I hear they’re trying to buy it back and restart with their uber gains

negotiation point for whom the recruiter or does the recruit actually know about the numbers?

The revenue has been sucked out of the publications pushing them into options like this. Craig's fortune was redirected from somewhere in a zero sum ecosystem. LLM searches making it far worse. I hope micropayments for original content work out but that's a pipe dream so far

Reader mode in mobile? But then maybe you lose some graphics

Richard bartle and the original MUD came out of Exeter uni AI program. Until this week he was a game design prof there too. See https://en.wikipedia.org/wiki/Bartle_taxonomy_of_player_type...

Learning prompting? Or how to set up your environment with a readme? I think they're are ways to be more productive with riding the vibe!

wow congratulations to you. I hope you weren't in a foreign country at the same time! I can understand why you might be a creature of habit and an unplanned bus routing is definitely a monkey wrench

How did you measure that? Podcast ads seem almost like old school TV in that a lot of listening is offline, with no call to action or method of measurement. Despite racing and resonating deeply with an audience.

Maybe have some off limits to glean shit posting channels?

Hi I've been working on something similar but using podcasts as a starting point for concepts. I wonder if we might collaborate? Link in bio, anon!

Does this play nice with react or similar Dom diffing frameworks? Can the animation values be fed in through react state variables?

Langchain did something overloading the | or operator in their LCEL DSL to form LLM pipelines. It feels like abusing python but it's familiar enough as a Unix syntax

A friend is building service called memxi that scrapes thr page you bookmark and then gives u a rag like way to look up things later with LLM fuzzy search

isn’t this just semantic chunking? There’s been white papers and already implementations in Langchain and llama index you could look through.

from what i see you have to guess a word from the list given

yellow squares are how many tries you have

as you put a word in, it will get added to the list of guesses

the closer words are at the bottom (thanks @ukv) so the goal is to extrapolate, eg is it an animal name, or an object? based on the order of your previous tries.

the font-size thing is a complete misnomer as its not really perceptible. but the insert location for guesses is more useful (but not documented)

there may be sort algorithm strategies for playing this, eg try two most distant words first... but the nature of the multi-dimensional word-distance being flattened to a single dimension makes it really hard to guess, imho.

when you get it right it deletes all word context and history.

interesting idea but a front-end dev could help make it playable!

https://scholar.google.co.in/citations?user=p6QKLIMAAAAJ&hl=...

does the bing API have citations? both you.com and pplx have that feature which chatGPT doesn't, tho its still in closed beta release for pplx

Awesome!

if the fediverse supports posting with markdown and code syntax highlighting, what is the benefit of a standalone gist thing? - Perhaps the ability to edit/clone and update? - the anonymous dumps use case (which this doesn't support yet)

I think Tokyo still has the same thing, technically

So you use it for web surfing?

Please post if you find anything!

update I guess cog files run fine on fly.

https://news.ycombinator.com/item?id=39411748

replicate is also very hard to predict costs on, I've found their salespeople are reluctant to make any predictions since things are changing so quickly. So it might take 5mins to cold-boot a model for a 2s prediction run, but its not clear how much you pay for that run.

Replicate created the cog spec and is a fantastic resource for browsing and playing with new models. They are a social destination too.

But flyIO is nice and simple for docker side projects, I hope their cog deployments are as smooth as there are a few extra pieces involved.

Any reason you're using rest for a chat client? Maybe socket scaling limitations of the node/express stack. Not as bad as php whci slack is still mostly written in, but hard to beat the mostly Erlang stack for discord. It also makes it's lot easier for bot dev having a two way connection without the hassle of ngrok etc.