viathebrink

prev @poolsideai
tapping life’s ley lines

[viathebrink @viathebrink]

And she claims it all came to her in a dream too - incredible powerful schizo aura

> [Empress] says it was a dream she had one night in 2014. Chains made up of numbers wrapped around the video game Dark Souls 2. It was anti-piracy software, she realized. As she focused, she began to see “what every number meant ‘universally,’’’ she says. Turning deeper inward, she says, she soon entered “the ‘ZONE,’ which allows me to SEE MORE into everything.” The chains broke.

[viathebrink @viathebrink]

@Nexuist snapcompact still getting roughly 2-3x effective usage out of context windows near losslessly, not exclusive to local agents
https://t.co/bpjbDt7uRv

[viathebrink @viathebrink]

@aaaronwill @alxfazio I need to pivot to omp lifestyle posting obviously

[yung algo @yungalgorithm]

check out pr #24 i opened to convert musicxml notation into notes so ppl can do fun arrangements and performances from musicxml files

[viathebrink @viathebrink]

@oleksoleksoleks

[viathebrink @viathebrink]

Need session trajectories or it’s bust, but depending on scale use a cheap-> hierarchy to process. Ex. Triage/classify with determinism or cheap model -> drill down ambiguous cases with model able to see the full trajectory

Don’t let things steer towards too much of a hard metric/rubric to avoid losing fidelity/nuance. Almost always going to be more valuable to scan with jev/haiku cost than to let things fall thru cracks with keyword/regex/similar filtering

After that it’s just allowing A/B experimentation based on findings, and having a solid(ish) replay system to be able to test changes

[viathebrink @viathebrink]

@wolfie_ Foot reveal 🦶

[viathebrink @viathebrink]

@aaaronwill

[viathebrink @viathebrink]

It assumes prior context is useful to hang onto which agreed isn’t always the case

Prompting outright in a way that’s less likely to land in an early basin works better too, if you know you’re going to have explore more solution space

Old pre-swarm prompt openai used on conjectures, but works really well if you adjust it to your specific problem

[Muratcan Koylan @muratcan]

The prompt engineering here is super impressive!

Such a great example of agent prompting:

[viathebrink @viathebrink]

@teortaxesTex https://t.co/tZIwwZctt6

[Catnostate @StringCatte]

@odoreida Immortality belongs to the butt breather.

[viathebrink @viathebrink]

@can1357 CHILL

[viathebrink @viathebrink]

@autofelatio2 @wolfie_ Wrist flex @wolfie_

[viathebrink @viathebrink]

@autofelatio2 @wolfie_ Jidion BONES

[viathebrink @viathebrink]

@PaulSkallas Chronically high dopamine levels tweaking -> lowered prolactin -> no hair loss 🤔

[viathebrink @viathebrink]

Your confusion is in taking the ToS at face value as opposed to the observed actions of the labs. Google also claims in ToS that only antigravity services should be disabled on misuse, yet they enforced full account deactivations

Understand that these are large orgs whose public facing commitments don’t always reflect their internally taken actions

[viathebrink @viathebrink]

A good start would actually be weighing whether what you post actually contributes meaningfully on a signal::noise ratio

Half the cua RFCs I got tagged on obviously had not been compared against the public cua roadmap, nor considered any code implemented as intentional. So the recommendations are completely off base and in-actionable

This is a result of you driving them completely with agents and never digging into the codebase yourself

The omp issue you opened regarding the ToS docs is another example. While correct on technicality, if you had spent any time digging through omp and its opinions, you’d know that NOT coupling the backend to an external sdk like agy is an intentional decision. So blindly opening the issue because you read the ToS and Claude told you they’re in disagreement is not a helpful contribution

[viathebrink @viathebrink]

@_kvnloo @michaelgrant @can1357 Also side note I’ve seen your kvnloo account opening a bunch of poorly calibrated RFCs, issues, and PRs across cua and oh my pi. They’ve tagged me quite a few times for poor reasons

Would recommend tightening whatever your setup is to be more intentional

[viathebrink @viathebrink]

@_kvnloo @michaelgrant @can1357 Don’t spread fud without personally testing

[viathebrink @viathebrink]

@michaelgrant @can1357 Have been running 2 anth max plans through omp for ~6mo without issues. Anyone telling you otherwise is fudding without testing themselves

[viathebrink @viathebrink]

@jtevesobs Vocals from this session were so raw 🔥🔥