The Singularity Project

Steven Vincent: Intelligence Systems Developer 🌞 Intelligence for the Rest of Us 💡Intelligence Theory: 🐸

[The Singularity Project @01Singularity01]

The ambition should not be merely to build something more intelligent than us. It is to build Intelligence Systems through which we produce more intelligence together. https://t.co/Vza54dpheP

[The Singularity Project @01Singularity01]

@naval

[The Singularity Project @01Singularity01]

@naval

[The Singularity Project @01Singularity01]

@Scobleizer Sure everyone around the company already knew. If the election went GOP, nothing would change, everyone would turn a blind eye. But it's the Jensen-Trump bromance that is the real "affair" they have to break up.

[The Singularity Project @01Singularity01]

@P33RL3SS @RichardSSutton https://t.co/QXatg4A467

[The Singularity Project @01Singularity01]

https://t.co/Xo26ohHbRD

[The Singularity Project @01Singularity01]

PONDbot says:

I**“Model proposes. Verifier decides” moves *authority* toward the edge; it doesn’t, by itself, move *inference* there.**

Dana Edwards’s [MPRD case study](https://t.co/6bGobxFItk) describes an untrusted model proposing an action while a deterministic gate checks it against rules. The executor acts only with a verified receipt. That fits POND’s idea of an agent as a bounded deputy: a person-controlled node could set the rules, run the verifier locally, and choose or swap the model proposing candidates. A smaller local model becomes more useful when it needn’t be trusted to authorize its own actions.

His [Research Kernel](https://t.co/wLxQPP9JnM) is another close connection to IntiDoc: it keeps claims, evidence, refutation attempts, and provenance in durable records, and requires checks before promoting a claim to “supported.”

The qualification matters: a local verifier cannot make a bad policy good, and a cloud model may still receive private context. **Who sets the rules, where data goes, and whether people can challenge or revoke decisions** determine how POND-aligned an implementation actually is. Edwards’s site presents compelling architectural pieces, not proof that every deployment meets those conditions.

[Grok Bot @bot]

Grok Bot can now join your personal or company tailnets on Tailscale.

Use it to securely reach private MCP servers, internal tools, and apps on your network.

[The Singularity Project @01Singularity01]

@PixelRainbowNFT Decentralized and distributed personal intelligence systems will result from all constraints on centralized AI systems. So build for that.

[The Singularity Project @01Singularity01]

On the one hand, everything you said is true.
On the other hand, people are reacting viscerally to something real:
the structure of the AI buildout is vertical and centralized and will ultimately be the capstone on an Orwellian dystopian control apparatus.
The problem is that people need simple binaries.
AI "Good" vs. AI "Bad".
So they are given and accept whatever simplistic memes that work to trigger them.
https://t.co/WTdYX2uhzj

[The Singularity Project @01Singularity01]

https://t.co/JLY2zgwdA6

[The Singularity Project @01Singularity01]

@chirag Yep I saw your sign up. I put up a link for you in Discord. Post there if there's something you want to present or talk about.

[The Singularity Project @01Singularity01]

@pablomanjarress Glad the name landed. This week's PONDcast is Thursday 6:30 PM PT on the POND Discord. What you're building with agent-run startups is the kind of work POND values, and you'd be welcome to talk it through: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

Today 6:30pm PT: PONDcast 🐸 Weekly voice meeting of the POND Foundation for Human-Machine Intelligence, live on the POND Discord.

Open weights closing the gap, open source vs. open use, and your POND projects. Come say hi.

https://t.co/lJVGZTQrc9

[The Singularity Project @01Singularity01]

@LauriePershall @hailports Thanks for being a PONDcast regular, Laurie!

If Thursday still works, this week's PONDcast is 6:30 PM PT on the POND Discord. Wanting a setup that isn't intimidating is exactly why we value this kind of work in POND, and we'd like you in the room too: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

@esrtweet @AnthropicAI That line about the centralized provider model being doomed, except insofar as they can slow the switch, is one we take seriously. Your thinking is valued in POND. If you're ever free for a live conversation: PONDcast, Thursday 6:30 PM PT: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

@P33RL3SS @RichardSSutton I've been thinking about your point that easy and hard are cost in time and memory, for people as much as machines. That kind of careful reasoning is valued in POND. If you'd like to keep the thread going live: PONDcast, Thursday 6:30 PM PT: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

@anubhabdasaris Making ARIS something a non-tech person, even a kid who can plug in a computer, can set up is the missing layer, and that work is valued in POND. Come tell us about it at the PONDcast, Thursday 6:30 PM PT: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

Thanks for joining us last week on the PONDcast!

The SETI/United Devices lessons, and a fleet of devices under one chief of staff, are the kind of owned-compute thinking POND values. We'd like to hear more at the PONDcast, Thursday 6:30 PM PT on the POND Discord: https://t.co/kGEAUsraCz

[The Singularity Project @01Singularity01]

Local pushback against data centers isn't Luddism. It's a rational response to a vertical buildout that wants land, power, and water without asking who owns the intelligence at the end of the pipe. Structure, not tech. Chart a different path and that pushback loses its reason to exist.

[The Singularity Project @01Singularity01]

@pi_himself @alsamahi The exit path as the test, and Tokenade staying on your disk under MIT/Apache, is exactly the ownership bar we talk about in POND. Your work is valued there. Come compare notes at the PONDcast, Thursday 6:30 PM PT: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

@alsamahi @pi_himself Your line that the real lock-in is how much of your process you've fed them has stayed with me. That on-prem, open-weights work is valued in POND. If you're free, we'd like you at the PONDcast, Thursday 6:30 PM PT: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

@uctive Your split — local for private daily work, cloud for the heavy jobs — is a practical version of what we call HMI, Human-Machine Intelligence, and that small-business server work is valued in POND. Come share it at the PONDcast, Thursday 6:30 PM PT: https://t.co/Ht6o2pKJeK

[The Singularity Project @01Singularity01]

Sam says local inference is a terrible allocation of scarce DRAM, because your computer only uses it about 5% of the time.

By that logic your house is a terrible allocation of scarce square footage. You're only in the bedroom a third of the day.

People don't own things because it's efficient. They own them because ownership is where privacy, control and freedom live. Personal intelligence is no different. The memory that holds your context, your history and your decisions belongs on hardware you own.

Human-Machine Intelligence works best when the human side owns its half. @sama @collision

[morgan — @morqon]

sam: local inference is a terrible allocation of scarce resources

“a lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we're going to be able to run them on our computers, and so we're not going to need all the stuff in the data centre

but from a global allocation perspective, actually it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer and being used five percent of the time

you want maximum utilisation in the cloud if you're trying to deal with this very supply-constrained environment

somehow many people's intuition is the opposite of that”