... if tests are effectively the source code, the system’s variety is capped by the test suite’s scope.

Relying on tests is an ecologically expensive process of high-entropy mutation until a test passes. It produces 'Frankencode' — logic that satisfies the rubric but lacks semantic grounding. 2/2

From an evolutionary perspective, the 'tests' argument confuses selection pressure with the generative substrate.

While tests are actuators that trigger revision, they are not the source of intelligence. Moreover, per the law of requisite variety ... 1/2

Yeah, it seems identity always has a few more dimensions!

Just remembered this blog post. Much shorter and an easier read, and totally on point to the social graph question here.

philipsheldrake.com/2022/03/we-n...

For anyone who really wants to deep dive on this:

generative-identity.org/human-identi...

Weirdly, Proton Mail’s search defaults to searching email titles only. But there’s a toggle to include the main body too. This search has to be client side of course because Proton itself has no visibility given its dedication to very secure encryption.

Proton’s strong encryption means mail is not available to generic mail clients (like Apple Mail, Outlook, Spark) on iOS the same way a normal IMAP account is.

Proton’s Bridge is only supported on desktop platforms, not on iOS, so iPhone clients cannot “pretend” Proton is a regular IMAP server.

A paper on designing “systems where prompts don't need to be perfect, because the system's structure does most of the work, and the AI is just filling in blanks that are hard to fill incorrectly.”

1-page exec summary you’ll be delighted to know 😁

recognitive.io/docs/superso...

😆

Just checking … you’re not actually recommending this, are you?

simpler is good

clearer is good

faster is good

but these are all possible without also making the rules WEAKER

On the topic of Billions we’re presented with the idea that reputation scoring will somehow save us from dystopia. OMG. Non-contextual, non-permissioned reputation scoring IS dystopian.

I appreciate the topic discussed is challenging, so source some third party expertise for goodness sake.

This article is not impressive.

On the topic of World: “While these fears are somewhat overblown” with absolutely no rationale presented for such an authoritative sounding conclusion. 🤦‍♂️

Supersoftware is software that is itself symbolic AI. It can reason about itself.

All domains share a reactive and reflective symbolic system. It's syntactically and semantically homoiconic just like human language.

Our draft white paper: recognitive.io/docs/Recogni...

You might say the semantic web is technical and totalizing whereas the meaningful web is contextual, plural, natural, human.

To be most useful, LLMs must share our world model(s), our contexts, our grounding. And we can help them do that in neurosymbolic synergy with our supersoftware.

Semantic in semweb contexts --> the precise definition of words and their relationships within a specific language system; the objective technical aspects.

Meaningful --> the overall significance and implications of something, often involving subjective (i.e. human) and contextual interpretation.