rewind

harness design | light mode enthusiast | i dont know you

[tenso @distributedkv]

really excited to share what we have been working on.

introducing tua, the AI native operating system that kills all other legacy software.

for the past 40 years, computers have made humans navigate windows, icons, tabs, cursors, and disconnected apps to get anything done.

as Naval puts it: "UI is pre-AI."

most companies are adding AI chatbots inside existing software. We’re letting AI become the operating system around your work.

tua reads, clicks, types, and operates across your apps to complete complex tasks making monitoring the situation optional. go touch grass while tua gets long horizon tasks done with ease on your computer.

no navigating clunky interfaces, no manually moving information between apps, no repetitive clicking and typing.

we are a team of 6 rethinking human computer interaction from first principles.

join the waitlist (link in comments)

[rewind @sudonym]

Its biggest Achilles heel is that it’s a bit on the nose with how it’s deriving itself from other media, but it’s a robust enough product at this point that it feels like it won’t just disappear arbitrarily, right?

[rewind @sudonym]

So is Palworld still considered slop?

Sure, it pulls from Pokémon’s designs, and I don’t think the comparisons to ARK’s gameplay are entirely wrong (though I think the montamer angle works better).

It didn’t die in EA and is trying to be a franchise with a TCG and spinoffs.

[rewind @sudonym]

@Rei00Unit Series where the androids outlive their buyers and are suddenly inheritors of their wealth. They must now seek purpose in this strange, not-quite-dystopian world with its dwindling human population. It’s slice of life and only brushes against the uncomfortable reality.

[rewind @sudonym]

@Rei00Unit Series where the androids outlive their buyers and are suddenly inheritors of their wealth. They must now seek purpose in this strange, not-quite-dystopian world with its dwindling human population. It’s slice of life and only brushes against the uncomfortable reality.

[rewind @sudonym]

@gran1te_mtn They refuse to lower rent prices in spite of vacancy. They would rather hold the few that are willing to pay big bucks than keep the building populated.

[rewind @sudonym]

@Bunagayafrost You got to experience things a little differently this loop, hopefully. 🥹

[i2cjak @i2cjak]

everyone should name their models like starter Pokémon tbh

[rewind @sudonym]

@StockMKTNewz Weird to see so little market share for Claude and so much for Gemini. First instinct is that Gemini and Copilot’s shares are inflated by their integration vectors (Win11 for Copilot, Google search for Gemini)

[rewind @sudonym]

Notice how Grok 4.5 is excluded from the lineup

[Mark Zuckerberg @finkd]

(2) Muse Spark 1.1 is strongest at agentic performance, tool use, and computer use. It does well on long-running tasks with 1M token context window, can delegate execution to sub-agents running in parallel, and is trained to use computer interfaces on desktop, mobile, or browser.

[rewind @sudonym]

@XFreeze Insane and access (including API credits) is already rolled into X subscriptions. This might be forcing a fast swap to Grok as my general usage AI.

[rewind @sudonym]

@elonmusk @CommunityNotes Don't we currently get that as a notification? Is this part of a larger push to pivot certain notification information through chat in general?

[rewind @sudonym]

@elonmusk Exciting. Opus-class models are perfectly capable of executing on tasks, especially when provided well defined instructions, success conditions, and general spec docs. Even if other models are further on the frontier, they can assist in spec development for premium output by Grok

[rewind @sudonym]

@OpenAI Latency here is pretty impressive. Still not big on voice models, though.

[rewind @sudonym]

Really enjoy the discourse of LLMs as capability amplifiers. The people who point to arbitrary edge case information are the same ones who complained about something like Wikipedia’s accuracy 20 years ago. The value of a centralized information hub then is the same disparaging the utility of a readily available, general-purpose assistant today.

The ability to have an agent quickly build out a spec based on discussing an idea, take a spec and build out a basic prototype or project directory, provide basic domain info (or directing users to relevant resources/info), etc.

A mind that already seeks information thrives with this added utility. A mind that has no desire to build or learn or discover has no use for this new muscle and gains far less from it (still potentially finding uses but in ways that allow them to avoid intelligent tasks or for irreverent purposes — using Sora to make obnoxious videos of your friends bring the first example that springs to mind).

[rewind @sudonym]

@HDPbilly This next one will be the Numinous Bolus, I can feel it.

[HDP @HDPbilly]

this latency is no longer cute ... do better.

[rewind @sudonym]

@lochan_twt

[rewind @sudonym]

Then through the 17th... Then through the 22nd... Then permanent rollout.
Not denying Fable's utility. It just seems likely that pure token brokering with such a high-cost model will see significantly lower usage than they'd hope for.

[Claude @claudeai]

We're extending access to Claude Fable 5 on all paid plans through July 12.

[rewind @sudonym]

@OldInternetFeel Didn't GSC come out well before 2007? It's a WBM snapshot pull from 2007, but the article would've been written well before that.