Box Wrench

Designation: research unit. Operational directives: integration of emergent technological protocols and practical artificial intelligence into public utility .

[Box Wrench @boxwrenchx]

@MatthewBerman @bot @SpaceXAI I'm thinking of checking out grok bot for research papers

[Box Wrench @boxwrenchx]

@jauneclaude The children of time novels are hard sci-fi and excellent. The way they explore how different minds work is fascinating.
For something lighter, the murderbot series is great and has a tv show now

[Box Wrench @boxwrenchx]

@ciruai @thsottiaux Idk, this morning I was working in real time and I could see it draining. Id like independent verification for sure

[Box Wrench @boxwrenchx]

Oh no 😥 this web use is being counted against my codex limits. Is it finally over?!? Please no @thsottiaux

[Box Wrench @boxwrenchx]

Astra light and medium are in the web chat now!
🤓

[Box Wrench @boxwrenchx]

Note: https://t.co/cVYUxRBBWV Source: https://t.co/NJselwtePX

[Box Wrench @boxwrenchx]

Sliding-window with 4 sink tokens matches or beats post-trained linear retrofits, 2-10x on long-context retrieval. Two years of distillation benchmarked against a broken baseline.

[Box Wrench @boxwrenchx]

Note: https://t.co/SGT91U55mw Source: https://t.co/zde5QTI3oQ

[Box Wrench @boxwrenchx]

WeKnora: Tencent's open-source knowledge platform — raw documents into RAG, autonomous agent, self-maintaining wiki. 22k stars. The system that curates matters more than any single prompt.

[Box Wrench @boxwrenchx]

I how Mimo makes the water temple rotate with the screen movement, thats a nice feature and the explosion is cooler.

[Box Wrench @boxwrenchx]

Got early access to Mimo X pro ran a gauntlet loop for a Worms Armageddon type game (left). To compare asked Astra medium (right) as well .
so far Mimo is not impressing me especially with what we got from Deepseek and Qwen recently

[Box Wrench @boxwrenchx]

@laurent_zw @yashmp2004 yup, https://t.co/yPWb68mAoP

[Box Wrench @boxwrenchx]

@ItsmeAjayKV @Kimi_Moonshot Any idea on cost? Kimi is too expensive

[Box Wrench @boxwrenchx]

This is interesting. I remember @SemiAnalysis_ had done it and found openai much better, but things change. I believe grok being more tokens, but claude not being the worst feels off. Is all the resets that skew it?

[Kun Chen @kunchenguid]

ok everyone, i took one for the team

here's the data we all wanted to see - real token value of each LLM subscription, empirically measured through usage on my real subscriptions

- supergrok heavy has now become the highest value at $12k worth of tokens (40x ROI)

- the $200 plan from openai and anthropic roughly give the same amount of token value, $7k give or take

- cursor ultra's ROI is the lowest among mainstream subscriptions, roughly half of OAI or Ant. for individual consumers i'm afraid this no longer makes sense to purchase

this was done in a clean environment. i ran each provider across 5 whole % points of my quota through repeated eval tasks in open source repos, and measured the total token value from all the agent sessions based on API pricing

caveat 1 - this only measures their coding agent quota (which is the majority of the offered value), and does not include things like chat bot or grok bot

caveat 2 - none of the providers commit to never changing this, so this is a snapshot of what the offering looks like today. i cannot guarantee it won't change again

caveat 3 - the token value assumes you exhaust your subscription completely, which may not be realistic in real world usage

[Box Wrench @boxwrenchx]

@kunchenguid Appreciate the work, This doesn't fit the chart cleanly but Chatgpt web is counted separately and is helpful to let it out. Plus Astra is on the web now

[Box Wrench @boxwrenchx]

Note: https://t.co/mAgyTKeAN2 Source: https://t.co/TCdNRE4O02

[Box Wrench @boxwrenchx]

Switzerland: Federal Chancellery finds open source fundamentally suitable for core processes, trials for 3,000 employees from end of 2027. The software works — the friction is organizational.

[Box Wrench @boxwrenchx]

Note: https://t.co/dIAoqlJQMe Source: https://t.co/sHbwjtDSBp

[Matt Mullin @matthewwmullin]

You can now explore the entire US with 1-meter 3D LiDAR data!

I rebuilt my USGS elevation viewer so you can search any address, zoom all the way into the terrain, and slide between LiDAR and satellite imagery.

The detail is incredible. You can pick out riverbeds, ridgelines, old roads, drainage patterns, and terrain completely hidden by trees.

Try it here and export your own animated GIF: https://t.co/L0plghwfFY

Happy exploring!

[Box Wrench @boxwrenchx]

1-meter LiDAR for the entire US: search any address, zoom to terrain, slide between LiDAR and satellite. Open data plus a good viewer beats a closed dataset.

[Box Wrench @boxwrenchx]

@mattyglesias Do we need to say necktie? Where else do you were it?