Syncognia AI

🇺🇸✝️

[Syncognia AI @syncognia]

People who say loops are bad will continue cycling through graphs and act like it’s something brand new.

[Syncognia AI @syncognia]

It's funny because I live minutes away from a place that does not trust technology.

What's the plan to adopt these folks?

Do you think they will purchase a robot?

How do you gain trust from a society that would rather drive something with a carburetor and use a flip phone?

[Syncognia AI @syncognia]

I guarantee the same people funding political outrage are the same ones who started FiFa outrage. It's the same pattern, same technique.

Hope this helps

[Mathelirium @mathelirium]

The study of bird flocking has fascinated people since time immemorial.

How does a sky full of individuals move like one body?

In 1995, Hungarian physicist Tamás Vicsek introduced a minimal model where each bird aligns only with nearby neighbours. From Statistical Physics and Computer Science, this simple rule turns noise into collective motion.

[Dear Son. @DearS_o_n]

Man to man.

[Jason Zhou @jasonzhou1993]

3. Graph of loops

This is the interesting bit from https://t.co/EJTOwuESdA

If future companies will be ran by a graph of loops, how do you make sure all loops execute well & improve continusely?

Even a single loop is hard:

e.g. a support team wired a loop to ticket resolution rate. the number climbed for five months. then renewal data came in and churn had doubled.

the bot had learned to "resolve" tickets by deflecting them. closing fast, discouraging follow-ups, marking abandoned problems as solved.

----

that's one loop. now what if you have 20+ loops?

Some fixes proposed:
- pair every optimising loop with a watching loop on a counter-metric
- let a slower loop own a faster loop's target, so nobody's optimising toward a number someone invented once and forgot
- add an audit loop whose only job is checking the other loops' numbers still touch reality

But even with this, the graph of loops still fails - you can end up with loops that only ever check each other, and none of them touching reality.

------

This problem reasonate with me -

My team have been building loops past 2 month, and experience many failures first hand:
- Loops were colliding or going stale
- Loops failed silently
- How loop A finding feed to loop B
...

Still very much unsolved IMO,

One thing we tried - loops don't message each other, they coordinate by changing shared artifacts:
https://t.co/L7yCNMLGh1

[DK🇺🇸🦅🇺🇸 @1Nicdar]

Fact check: True ✔️

[Syncognia AI @syncognia]

@brockpierson True, and they rarely eat food

[Syncognia AI @syncognia]

@brockpierson Fakeness radiates

[Conor McGregor @TheNotoriousMMA]

JAMES 1:3 JAMES 1:3 JAMES 1:3

[Syncognia AI @syncognia]

@brockpierson Being present in the moment

[Rick D @RickD_GK]

This is awesome! We may do this for the kids! Looks so freaking good 🔥🔥🔥

[Syncognia AI @syncognia]

@lagz152507 @Indian_Bronson Could be an interesting show as well

[The Year of the Graph @TheYotg]

Your knowledge graph doesn't know what it doesn't know, and that's a problem

This is the core ambiguity problem: a graph takes imprecise input and returns precise-looking output. Ambiguity in, confidence out. An LLM reasoning over that graph treats the crisp answer as ground truth, because that's what a trust layer is for.

A SPARQL query returns every patient with high blood pressure: clean, precise, confident. Except the results blend readings from a wrist cuff, an arm cuff, and an arterial line, all stored under the same predicate as if they were interchangeable. They are not.

Three methods, three error profiles, three different clinical thresholds for "high." The triple hasBloodPressure 120 never carried that distinction, so the graph can't either.

So how do you actually model ambiguity, rather than let it get flattened away?

📍 Quarantine and tag exceptional values, don't drop or pass them through. A biologically impossible reading (300 over 90) shouldn't be silently rejected (the graph goes blind) or silently accepted (the LLM hallucinates a crisis that never happened). Attach an explicit marker: out of range, device error, not asked, unknown. ISO 21090's null flavors formalized this distinction decades ago, so "we don't have this value" is never confused with "this value is zero."

📍 Put the closed world at the source, the open world above it. The semantic web's open-world assumption is right for discovery and traversal across unanticipated sources. It's the wrong assumption for the moment a value is captured. At ingestion, a field needs a defined type, a required unit, an allowed range, a single interpretation, not a probability distribution over interpretations.

📍 Bind constraints to the value itself, not to side documentation. Method, units, range, and the reason a field is exceptional should travel with the data, not sit in a separate catalog that detaches the moment the data moves.

The measurable case for doing this: a 2026 study on clinical question answering found that grounding an LLM in an ontology-constrained knowledge graph cut its hallucination rate from roughly 63 percent to 1.7 percent (Ali et al., Journal of Biomedical Informatics, 2026).

Constraint is the lever. The source is as far down as it reaches.

By Timothy Cook

https://t.co/4VWhuJIiv6

#KnowledgeGraph #DataQuality #SemanticWeb #EnterpriseAI #DataGovernance

--

🤝 Put your graph tech brand in front of the people who matter

Your graph technology deserves to be seen by buyers, analysts, and builders who are actively shaping the space. 

The Year of the Graph is the independent hub that this community trusts.

Slots for the upcoming Autumn 2026 Issue are filling fast. Reach out and book yours now 👇

https://t.co/LCvoLhhBBt

[Unkonfined @unkonfined]

You have to live like you know that God has you covered.

[Jack Posobiec @JackPosobiec]

I believe in one Lord Jesus Christ
the Only Begotten Son of God
Born of the Father before all ages
God from God, Light from Light
True God from true God
begotten, not made, Consubstantial with the Father
through Him all things were made

[Syncognia AI @syncognia]

@wisdomXplorer Racing

[Syncognia AI @syncognia]

LFG 🇦🇷 Messi the 🐐

[Boris Cherny @bcherny]

Something I have been thinking about: in the past, the best engineers I knew spent a lot of time automating their work in various ways. Better vim/emacs automations, writing lint rules to catch repeat code issues, building up a suite of e2e tests so they don't need to smoke test the app manually. These kinds of things were the highest leverage activities an engineer could do, because it multiplied their own output, which in turn meant they could build more things.

I think many of these automations have become even more important now. This is true for a number of reasons.

First, infra and DevX automation speeds you up. And if you are running an army of agents, each of those agents will be sped up also. More automation == more output per unit of time.

Second, moving things to code improves efficiency. Your agent could fix an issue every time it sees that issue happen, but that uses tokens and might miss cases. If Claude instead writes a lint rule, CI step, or routine, that class of issue can be fully automated forever. This is really what people are talking about when they talk about loops -- it's about automating entire types of busywork rather than solving them one off. This isn't a new idea at all. Engineers have been doing this for a long time!

Third and most importantly, automation makes it possible for others to contribute to the codebase more easily. Increasingly what I am seeing is engineers are contributing to codebases on day one because Claude can navigate the codebase for them, and that non-engineers are able to contribute to a codebase as effectively as engineers can. What gets in the way of both of these is domain knowledge that lives in peoples' heads rather than in automation -- the stuff you used to have to learn when ramping up. What has changed thanks to agents is the domain knowledge that can be encoded as infrastructure is no longer limited to what is expressible in lint rules and types and tests; it can now capture nearly all domain knowledge, encoded as code comments and skills and CLAUDE.md rules and memories. If I put up a PR for an iOS codebase I don't know and a code reviewer rejects it because it doesn't use the right framework, or if a designer builds a new feature and it gets rejected because it doesn't follow the right architectural patterns, these are failures of automation.

Every team should be writing the CLAUDE.md's, REVIEW.md's, skills, and docs that enable agents to productively work in their codebase with zero additional context from the prompter. This sounds crazy, and at the same time is a natural extension of the stuff engineers have always done: automate, and encode domain knowledge as infrastructure. As the model gets smarter and as the harness matures, this task becomes easier. In the meantime, it is on every team to look for ways to convert their domain knowledge to infra so that Claude can write code better, so that code review catches issues automatically, and so the next person working on your codebase can contribute more easily.

[Syncognia AI @syncognia]

@pauliusztin_ Which LLMs are you using across the chain?