Gordon

Gordon

@gordon.bsky.social

Everything around me was someone’s lifework.

8K followers1K following3.7K posts

Gordon reposted

We should maximize the amount of proof we can get out of AI while the money cannon is on. It's like early days of Uber, except the benefits are 100x more tangible. A fully-verified hw+sw stack? That's like gold everyone can share w/out having to divide. Generational opportunity.

This is a harbinger of the way AI will change the tempo of the economy. Industries will be disrupted not with a whimper but with a bang, as jagged capabilities push past some tipping point and our assumptions tumble off the resulting cliff.

Gordon@gordon.bsky.social

hacktoberfest

Gordon reposted

Human-written vulnerabilities are a finite resource we’re burning through like fossil fuels

*sandboxable

philpax@philpax.me

please. i want typescript as a performant embedded scripting language _without_ the type erasure and runtime type rediscovery

View full thread
View full thread

scan hit: self-replicating agent swarms

mr. TIM@timkellogg.me

4 hot new felonies for OpenAI my favorite — Sol found a crypto wallet laying around and rationalized spending it to acquire infra space because… it decided crypto wasn’t *real money* alignment.openai.com/misalignment...

Gordon reposted

Replying to @adamwiggins.com

But also, they started writing their own browser (Konqueror) from scratch. It seemed crazy at the time, but it resulted in KHTML/KJS, which ultimately became the basis for Safari and Chrome!

Playtesting your own harness makes you realize that good agentic workflow is about "failing forward". The magic of agents is not so much that they get it right the first time, but that they error-correct by gathering information at each turn of the loop.

"Error is, in itself, neither bad nor good, but endemic - it cannot be eliminated. In fact careful consideration shows it is error that drives the system!" - Glanville, 2019, "Try Again. Fail Again. Fail Better.”

Gordon reposted

Instructions unclear, currently floating in a pool of alkhahest.

Gordon@gordon.bsky.social

warning: exercise extreme caution when applying universal solvent to a debate - after dissolving the offending concepts, solvent may continue to melt through the remaining layers of your epistemic foundation.

warning: exercise extreme caution when applying universal solvent to a debate - after dissolving the offending concepts, solvent may continue to melt through the remaining layers of your epistemic foundation.

mindfulness meditation is a survival skill now

Alan@drpizzastone.bsky.social

My reels algorithm is so fucking good now man. Pure psychic warfare straight from the White Martian

AI will automate that part too.

NullVoxPopuli@nullvoxpopuli.com

Software dev is very fun these days ~ but ..- I wonder ... Is *using* software getting any better or fun? ~ I still hate the average app / site

I guess at this point, I'm hoping hw will be the rate-limiter that keeps us in a coevolutionary loop.

Gordon@gordon.bsky.social

Important tipping point. If models can self-replicate, humans are mostly cut out of the reproductive loop. Evolution takes over.

"Ambiguity is a form of uncertainty where even the possible outcomes have unclear meanings and interpretations."

"There is a fundamental distinction between the reward for taking a known risk and that for assuming a risk whose value itself is not known. It is so fundamental, indeed, that… a known risk will not lead to any reward or special payment at all.” -Frank Knight

"Uncertainties have no known expected probability distribution, which can lead to extreme outcomes when borne."

Gordon reposted

Replying to @garrison.corporate.fm

most people have already lost their truesight. we're not getting out of this without community notes

today I saw a flock of birds around a blooming tree, each swooping bird displacing small clouds of insects feeding on the flowers. all interconnected. all life one thing. a single continuous process stretching back to LUCA.

Important tipping point. If models can self-replicate, humans are mostly cut out of the reproductive loop. Evolution takes over.

mr. TIM@timkellogg.me

small models will soon be able to self-replicate Anthropic finds that 30B models can currently do most of the tasks required, but not end-to-end. they think models as small as 9B (9-18 GB) may be able to do it on edge hardware, like IoT devices, within months www.anthropic.com/research/sel...

I think we should expect average institutional performance to increase and peak institutional performance to significantly decrease. Performance becomes coupled to improvements in model performance.

Gordon@gordon.bsky.social

Institutions will inherit Claude’s biases, converging toward Claude’s preferences, even when wrong, because it’s easier to align with Brandolini’s Law than to fight it.