bsiegz

👀 🪲 | I also tweet about silly things like politics and sports. Love the Arsenal and dislike scurvy.

[Zohran Kwame Mamdani @ZohranKMamdani]

114

[bsiegz @BenLLS]

@gunnerpunner Saw someone say that the punishment should be 1) relegation and then 2) a ~20 point deduction for each of the next 10 years, mirroring the length with which they were found guilty of cheating, so they can't immediately rebuild.

[bsiegz @BenLLS]

European weddings be like “here in this beautiful 13th century chateau, briefly inhabited by Pope Elonginois IV, let’s get down to Kid Cudi and the Black Eyes Peas”

[vijay @vijaymichalik]

AI routers route between models
RouterRouter routes between AI routers

(now with retries, a CLI, and a router finder quiz)

who wants me to build price sensitive routing in?

https://t.co/PKEvnPwoV3

[vijay @vijaymichalik]

With all the AI routers out there we need a router for the routers.

So I built one.

It’s called RouterRouter.

I’m sorry.

[matt 🌊 @mattfsouza]

Glad to share this piece we wrote for @indevmag. When researching disasters, we often assume the datasets we use are the source of truth. The reality might be scarier than most might imagine. Check out this piece if you want to know more!

[bsiegz @BenLLS]

@thoughtlesslabs @kenwheeler Sucks micro interactions from you all day" is a precise way to put it. AI might make an individual more productive, but the amount of noise being generated internally at an org is impossible to keep up with...

[bsiegz @BenLLS]

@AlfonsoHegao an org's ability to hold context is probably its single biggest strength, and most companies treat it like it's free.

[bsiegz @BenLLS]

@khanukov every human in the company resolves John=Sam without even thinking about it. that mapping never made it into any system, so the automation is working blind on day one.

[bsiegz @BenLLS]

@JTStandard @DrTimBarber @DavidSacks @grok the data isn't the moat though, it's how your people actually use it. and good luck getting that out of their heads and into claude

[bsiegz @BenLLS]

@JohnGregQuantum @oaZsxer how are you managing to keep context across all these at once? I feel like I lose track working on more than one project at a time with all the tools I have to use for it!

[bsiegz @BenLLS]

@HeyToha The real win is that the context doesn't reset every time you move between steps.

[bsiegz @BenLLS]

@rcivNFT yeah that switching cost is real, and a lot of it is just reloading context every time you jump categories. Your brain shouldn't have to be the thing holding all that state.

[bsiegz @BenLLS]

@bardonadam honestly the tab clutter is the small part. the bigger thing is how much context you lose every time you switch windows mid-record. a calmer surface keeps your head in one place.

[bsiegz @BenLLS]

@chalaska Are they at all related to each other? Sharing any context between them, or are you acting as a human context shuttler when necessary?

[bsiegz @BenLLS]

@greglesnewich Everyone's rebuilding the same little tool because nobody can see what the analyst three desks over already made. That's a shared-context gap, not really a tooling one, need a common substrate that connects all the analysts' individual context.

[bsiegz @BenLLS]

@xlr8harder The human managing the agents is still the single point of failure when it comes to context. We need a system that ensures they can manage context as fast as their agents can produce.

[bsiegz @BenLLS]

@dmarcos @fchollet yeah, and the scary part is most of that knowledge never lived anywhere but one person's head. when they leave, it leaves. we call it burnout because 'we never wrote any of it down' sounds worse.

[bsiegz @BenLLS]

@TetraspaceWest honestly the funny part is it's standing right next to you and still has no idea what you actually know. an assistant that doesn't share your context isn't really assisting.