Brittle if/then/else logic gives way to robust or even anti-fragile logic with predicates.
github.com/monetic-labs/an-tâŚ
Gonna aim for this thing to be personalization of your information flow by default, bots will be native buried to a type of verified account. Youâll be able to control privacy across threads individually, sometimes you want/need/forced to be public, do so while in control of that blast radius
notdevin retweeted
I'm still rawdogging codex, claude, and others in TUIs over tmux, and one recurring painpoint is getting them to coordinate properly, so I built something that I want to share!
It's called flopwire: agent messaging and trace search that works across models, harnesses, machines/VMs/clouds, and teammates. It's called flopwire!
Here is Codex asking a live Claude Code session what it is working on, reading its transcript, and splitting the work.
flopwire.com
github.com/flopwire/flopwireâŚ
Why wouldnât your harness have dual environment configurations for your custom email service so you can run a full featured test suite and prototype out all your agent capabilities?
And not some wrapper over Gmail, a million LOC to handle spam by rules is probably not the best option in 2026, just sayin
I have 99% of the infra needed to build out farcaster competitors. Since itâs all open and free, I would like to build a network that causes everyone in the network state to groan in the morning when they think about the existence of this app. Ideally the app will be so effective at making them hate it that they abandon their client and have to start all over again (I have no idea what this means in practice)
We arenât aiming at anything other than anarchy with this, like goat simulator but for a network client and to make Balaji feel like we peed in his coffee
Uhhh, maybe? Donât most people now think of Gates as the dude that raped and ate kids with his boy Epstein on an island?
Either that or a backer to the pandemic who should be on trial for crimes against humanity.
Ben Horowitz: âThe most dangerous thing going on in AI is Anthropic and Bill Gatesâ
âAnybody can go say, âAI is going to kill us,â and you'll be on every major news station and every newspaper and so forth.
I think it's extremely dangerous, though, when it comes from Bill Gates because he is so credible, because he did build Microsoft, and people think of him that way, and then certainly he's been a big political donor, so he has a lot of political juice, particularly on the Democratic side.
The big risk that everybody talks about is that the models are going to recursively self-improve, and then they're going to take off, and then they're going to become so smart and run away from us.
Well, what would be the preconditions for that?
Ali Ghodsi from Databricks had a great formulation of, âWell, what would be the preconditions for that?â
And he had four preconditions.
One is that the next model has to require fewer GPUs than the last.
The second thing is the next model has to take less time to train.
And then the accuracy of the next model has to be better than the last, so it's got to be a better model.
And then finally, it's gotta keep happening again and again and again.
Zero of those four have been met.
We're not on track for any of those four to be met, and so we're really talking about a risk that's entirely theoretical at this point that we're not mapping to.â
If your company uses PRs and branches youâre going to be destroyed on pace in the market. Destroyed is too soft, your product will be irrelevant.
Some big companies will keep thinking they can buy their way in but every time they acquire they will reset under their releases rules which will be the equivalent of hitting the brakes on a plane, out of the sky you go
A common trope in tech is that you shouldnât do multiple things, some will even say you canât. Those people tell you this because they canât. They donât know how to do more than one thing at a time, theyâre identifying that their brains donât work that well and thatâs important because all of the most useful things to build are all behind layers of varied domains and domain expertise.
Capital alone will have very little power going into tomorrow because capital without the brain power to apply it in a novel way is going to be meaningless. Everyone with any capital will have yield applied to all idle capital without no cost to usability. The friction in the systems of money are going to vaporize and when it does inefficient capital becomes nothing but a liability and if the inefficient capital is managed by someone who canât figure out how to think about more than just one thing at a time then my guess is some system will come alleviate them of that money rapidly.
We look out and see stable enterprises and mountains of capital and think those last forever, they donât. We are in a very unstable societal configuration right now, the permanence of what we see is a mirage. Covid shut down an entire planet overnight, and you still believe in the fortitude of the institutions you see, your eyes deceive you. The world can flip overnight and most of what you see is already a shadow in the eyes of tomorrow
While the fireworks go off about 2023 defi and accounts over in big tech land, the future of accounts and money will be agent first self custodial designs that use DIDs, permissions as grants, and TSS to safely share and use credentials at the core of your service architecture all the way up to forward operations creating and interacting with external services
These accounts treat data you are responsible for as securely as we treat money. In a world with agents, novel data you own and generate is valuable
How do I bathe myself in this tweet? it feels so cleansing and refreshing
No it's not surprising. There's no such thing as CoT. It's fucking tokens that happen to follow one more "beginning of an italicised ~thinking~ block of text" than the number of "end of italicised ~thinking~ block of text" tokens.
The weight of all of these grown ass adult men fully having psychotic episodes in public every day on this shitty website is unbearable.
Oh now NOW THE MODEL IS SHOUTING BECAUSE IT'S ALL CAPS.
LMAO!
NOW IT'S LAUGHING IT'S ASS OFF? IT HAS AN ASS? WHAT IN THE WORLD MODEL?!!??
You people make me fucking sick. Get a real job you idiot grifters.
Skill issue, obviously
A harness like cursor doesnât appear to scale beyond ~3m LOC (rust). Context is not able to stay with what is needed, every prompt start to require manually connecting certain contexts. Maybe I need 10m context or maybe the harness doesnât get involved enough in the context passing to pull it forward correctly
A harness like cursor doesnât appear to scale beyond ~3m LOC (rust). Context is not able to stay with what is needed, every prompt start to require manually connecting certain contexts. Maybe I need 10m context or maybe the harness doesnât get involved enough in the context passing to pull it forward correctly
Sudden vision of the future where local bots bully the absolute fuck out of humans as obnoxious as Dario and Eliezer in order to demonstrate how not to behave in the world.
Our society has gone soft without bullying. There will be a short period of chaos as itâs reintroduced, we will lose a few blue hairs but they already opted out of procreation so nobody will actually care. Once the period of cleansing is over, the bullying will resolve to its minimal required state of casual mild berating and belittling











