Vernon Niven

COO of | marketing | investing | entrepreneur-ing
**all posts & retweets are my own**

[Vernon Niven @vniven]

@AlfieJCarter Agents

[Vernon Niven @vniven]

@Budowich Show us your proof.

[Vernon Niven @vniven]

@ronbodkin @JeffLadish For over two years, Anthropic has been calling for world governments to implement AI model regulation.

They are not the problem - they just don't want to use their tech for killing.

Perhaps that is a naive POV.

But that is something every CEO should decide - not the US govt.

[Vernon Niven @vniven]

@MilkRoadAI @DavidSacks Has anyone taken the time to look at David Sacks’ own hypocrisy on critical national issues?

Maybe we should understand this first, before listening to him?

(Just a thought)

Here are 5 good examples - courtesy of Claude ;-):

https://t.co/ikTCm8FNZL

[Suryansh Tiwari @Suryanshti777]

RAG might already be becoming obsolete.

A month ago, Andrej Karpathy dropped a simple GitHub gist called “LLM Wiki.”

Now the comments section looks like the birth of an entirely new AI category.

5000+ stars later, developers are rapidly building:
• persistent AI memory systems
• self-maintaining knowledge bases
• multi-agent research environments
• contradiction detection engines
• AI-native company operating systems
• local-first memory architectures
• graph-based reasoning layers
• evolving second brains

And the craziest part?

Most of them were built in DAYS.

Because the core idea is insanely powerful:

Instead of AI repeatedly retrieving raw chunks like traditional RAG…

…the model continuously maintains a living knowledge system.

Not temporary context.

Persistent synthesis.

The shift sounds subtle until you realize what it changes:

RAG:
retrieve → answer → forget

LLM Wiki:
ingest → synthesize → evolve

That one architectural difference is causing an explosion of experimentation right now.

People are already building:
• agent memory operating systems
• AI-maintained engineering documentation
• self-healing knowledge graphs
• persistent research environments
• conversational memory architectures
• contradiction-aware wikis
• context compression engines
• machine-readable company systems

The comments section alone feels like watching an ecosystem form in real time.

One developer built deterministic contradiction detection using sheaf cohomology

Another built “sleep consolidation” for AI memory systems inspired by human memory formation

Another created persistent multi-agent vault conversations

Another turned entire repositories into continuously maintained AI wikis

Another built local-first memory systems with audit trails, provenance, graph exports, and MCP integration

This is the important part:

Karpathy didn’t launch a product.

He introduced a pattern.

And patterns are what create ecosystems.

The same way:
• transformers created modern AI
• RAG created AI retrieval startups
• agents created orchestration frameworks

LLM Wikis may create persistent AI memory infrastructure.

That’s why this moment feels different.

For years, AI systems have been stateless.

Now developers are trying to build systems that actually accumulate understanding over time.

And once knowledge compounds instead of resetting…

…the entire interface layer of AI changes.

(Link in comments)

[Mira Murati @miramurati]

Today we're sharing our work on interaction models. A new class of model trained from scratch to handle real-time interaction natively, instead of gluing it onto a turn-based one.
https://t.co/MoS5s4cm60

[Andrej Karpathy @karpathy]

This works really well btw, at the end of your query ask your LLM to "structure your response as HTML", then view the generated file in your browser. I've also had some success asking the LLM to present its output as slideshows, etc.

More generally, imo audio is the human-preferred input to AIs but vision (images/animations/video) is the preferred output from them. Around a ~third of our brains are a massively parallel processor dedicated to vision, it is the 10-lane superhighway of information into brain. As AI improves, I think we'll see a progression that takes advantage:

1) raw text (hard/effortful to read)
2) markdown (bold, italic, headings, tables, a bit easier on the eyes) <-- current default
3) HTML (still procedural with underlying code, but a lot more flexibility on the graphics, layout, even interactivity) <-- early but forming new good default
...4,5,6,...
n) interactive neural videos/simulations

Imo the extrapolation (though the technology doesn't exist just yet) ends in some kind of interactive videos generated directly by a diffusion neural net. Many open questions as to how exact/procedural "Software 1.0" artifacts (e.g. interactive simulations) may be woven together with neural artifacts (diffusion grids), but generally something in the direction of the recently viral https://t.co/z21CP5iQfu

There are also improvements necessary and pending at the input. Audio nor text nor video alone are not enough, e.g. I feel a need to point/gesture to things on the screen, similar to all the things you would do with a person physically next to you and your computer screen.

TLDR The input/output mind meld between humans and AIs is ongoing and there is a lot of work to do and significant progress to be made, way before jumping all the way into neuralink-esque BCIs and all that. For what's worth exploring at the current stage, hot tip try ask for HTML.

[Vernon Niven @vniven]

Good, skeptical critique of Palantir/Karp’s manifesto.

[Mehdi (e/λ) @BetterCallMedhi]

I just finished reading palantir’s manifesto & I need you to understand what you’re actually looking at because this is the MOST important document the tech world has produced this year

most people came away thinking «wow what a thoughtful essay about patriotism and technology »…I came away thinking this is the most elegant justification for corporate capture of the state apparatus ever written & I want to walk you through why

krp opens with «silicon valley owes a moral debt to the country that made its rise possible » & frames the entire document as a call to civic duty, but read between the lines and what he’s actually saying is that the engineering elite should be embedded inside the defense and intelligence apparatus of the nation, he’s describing exactly what palantir has already done and dressing it up as patriotism

«the question is not whether AI weapons will be built, it is who will build them and for what purpose »sounds like a warning but it’s actually a sales pitch, he’s telling every gov on earth that the choice is binary either you buy from us or your adversaries will build it without you, this is the oldest arms dealer rhetoric in history wrapped in SV vocabulary

« hard power in this century will be built on software »is the key sentence of the entire manifesto because this is where karp reveals the real thesis, he’s saying whoever controls the software layer of national defense controls the nation itself & if you’ve been following my threads you know that palantir’s gotham and foundry platforms are already plugged into the intelligence feeds the satellite data, financial transactions & communications of dozens of govts worldwide through a single ontological knowledge graph that creates a technological dependency so deep that migrating away would mean rebuilding the entire institutional memory of the organization from scratch

this is vendor lockin at the scale of nation states and I’m personally convinced it was designed this way from the beginning

«we should applaud those who attempt to build where the market has failed to act » is karp defending palantir’s expansion into every domain the gov used to handle itself, policing immigration, military targeting intelligence analysis public health, everywhere the state retreats palantir advances and what was once a government function becomes a private service that the government can no longer perform without plantir’s permission

and here’s what I think makes it even more concerning, these systems are increasingly autonomous meaning the AI layer is making targeting recommendations threat assessments & resource allocation decisions that humans inside gov are rubber stamping without fully understanding the underlying logic

a bureaucrat inside the pentagon / DGSI sees a recommendation from the system & approves it because the system has been right 97% of the time and questioning it would require technical expertise that no one in the room has, this is algorithmic governance wearing the mask of human decision making

«the atomic age is ending, a new era of deterrence built on ai is set to begin »is the MOST chilling sentence in the document because karp is explicitly saying that ai based deterrence will replace nuclear deterrence as the organizing principle of global power, and whoever builds that ai deterrence layer owns the 21st century the same way whoever built the bomb owned the 20th & he’s telling you plainly that palantir intends to be that builder

«national service should be a universal duty » & « we should only fight the next war if everyone shares in the risk »sounds noble until you realize that he is proposing a system where citizens serve the state & the state is operationally dependent on palantir, the public bears the risk and palantir captures the value, soldiers fight wars planned by algorithms they can’t audit built by a company they can’t vote out

[Vernon Niven @vniven]

@MrKrueger70 @OfTheBraveUSA Here ya go:

https://t.co/4UQ7c9O8Dp

[Vernon Niven @vniven]

@delilah1378510 @SenRandPaul @Sec_Noem Um...

DOGE did not deliver anything close to a meaningful amount of savings - and the courts are reversing what's left.

Just FYI - I've got no dog in this hunt.

(unaffiliated voter)

[Vernon Niven @vniven]

@ISephara or...

The plane flew to Berlin to protect the aircraft from Iranian missile strikes.

[Vernon Niven @vniven]

@Ludwig64747662 @Cheeto_Chicken @stannny71 @krassenstein You don’t understand our electoral system.

Trump won in 2020 with a minority of votes, as well.

[Vernon Niven @vniven]

@Belindahenry18 @TxgrlThatsme @AstroNoodleCat @DHSgov DHS misspoke: you can't be "paroled" from an indictment.

Mr. Paktiawal was indicted in Dallas for 3rd-degree theft and released on a $3k bond; there is no record of a conviction for anything.

And: he did serve with US Special Forces.

https://t.co/QWTEFIR108

[Vernon Niven @vniven]

@HedgieMarkets Excellent find.

I just asked Claude Code to go through my personal libraries and scan for invisible Unicode using Python byte-level scanning.

Found none.

So I asked Claude to add a Unicode scanner to my platform to scrub every push.

Took all of 2 minutes.

[Vernon Niven @vniven]

@PrdAmerican1 @allenanalysis It is SO ironic that you post that comment here trying to be cutesy, on a platform specifically designed to amplify people who freak the f*ck out over anything.

Welcome to the club?

[Vernon Niven @vniven]

@chamath Chamath,

Republicans in several states have also pushed to ban some of these books and other masterpieces of literature from our schools for the same stupid reasons.

Snowflakes to the left, snowflakes to the right… let’s get back to the middle again 🎶🎼🎶

[Vernon Niven @vniven]

@asparagoid Wanna actually help people?

Then share your gold-standard / double-blind clinical evidence proving your case.

Just like the pharma companies did to get GLP-1 approved.

We don’t need another “research paper”.

Why should anyone listen to this pseudoscience BS, until you do?

[rvivek @rvivek]

An engineer at Anthropic wrote a spec, pointed Claude at an Asana board, and went home. Claude broke the spec into tickets, spawned agents for each one, and they started building independently.

When the agent is confused it runs git-blame and messages the right engineers in Slack. By Monday the agents finished the plugin feature.

That's one example of how the best engineers are shipping software right now.

Developers will soon orchestrate 50 AI agents in parallel and the difference between a good engineer & a great one would come down to specs.

You can't write a spec that holds up at that scale without genuinely understanding what you're building at a deeper level.

The next-gen developer who understands the fundamentals, can architect well and orchestrate agent is going to be a 1000x developer!

[Vernon Niven @vniven]

@pgavigan77 @WalshFreedom You obviously have a lot of learning to do. I wish you the best of luck.

[Vernon Niven @vniven]

@Fighter569063 @LP808bb @BillAckman (cough)