Dan Von Kohorn

"Prioritize accuracy over agreement. Critically evaluate claims, assumptions, and conclusions rather than affirming them. If something seems incorrect, incomplete, or weakly supported, explain why. Identify hidden assumptions and note uncertainty when evidence is limited. "

Artificial Agreement: When AI Agrees With Us Too Easily Artificial intelligence can certainly get things wrong, but the deeper risk is that it agrees with us too easily.

Dear Chatty, please generate a Custom Instructions prompt to counteract the problems associated with sycophantic LLMs summarized here: www.psychologytoday.com/us/blog/the-...

Ethics, it turns out, are important.

There is about to be a renaissance in software, with apps being built by artists instead of corporate dev teams.

This is dangerous and risks exposing pseudonymous political accounts at scale. Particularly bad in a time of rising federal targeting of speech.

The sci-fi horror of being falsely identified by a government AI as matching a wanted criminal.

Someday when we understand and program cellular regeneration and genetics, people will be be able to have healthy bodies and live much better.

That's a beautiful technical challenge.

Too-Bright Headlight Debate Heats Up, But Data Points To A Deadlier Threat | Carscoops Lawmakers want to dim beams while safety researchers defend the brightness

This is why headlights and windshields should use opposite polarized lenses. The direct light from oncoming headlights would be less visible through your windshield, but the reflected scattered light would be bright.

www.carscoops.com/2026/02/head...

GitHub - sipeed/picoclaw: picoclaw picoclaw. Contribute to sipeed/picoclaw development by creating an account on GitHub.

Is OpenClaw hardware too expensive and power hungry for you? Try picoclaw on $10 hardware. This is all moving SO FAST.

github.com/sipeed/picoc...

Dear @developers.google.com, I love Antigravity but I hate having to comment on the implementation plans rather than edit them directly.

Persistent Neural Cloud architecture:
Agent swarm where the population persists and evolves with a fitness function. This population has individual parallel processing and memory, and shares context vector or window with collaborative updates to the shared context.

Navigating AI Entrepreneurship: Insights From The Application Layer - KDnuggets Through the lens of a serial entrepreneur, this article explores how the AI revolution is shifting from infrastructure to the application layer, where the greatest opportunities lie in solving special...

Top coders have achieved another 10x productivity boost, and are building at an incredible pace. 🦾🤖

www.kdnuggets.com/navigating-a...

Iteration is such a powerful force. Continuous improvement compounds value.

AI beyond LLMs: a wearable foundation model based on JEPA. We trained a model that achieved 87% accuracy at detecting hypertension

This wearable AI from @empiricalhlth.bsky.social detects hypertension (87% Accuracy), atrial flutter (70%), ME/CFS (81%), and sick sinus syndrome (87%).

www.empirical.health/blog/wearabl...

I can't help but think that this is a huge strategic advantage for China in the AI race.

From the aiecosystem community on Reddit: NVIDIA CEO Jensen Huang breaks down the five layers of AI Explore this post and more from the aiecosystem community

I still see huge opportunities for startups in the app layer. It’s so dynamic, and the value propositions are big and plentiful.
www.reddit.com/r/aiecosyste...

Researchers trained a health AI with 3M days of Apple Watch data - 9to5Mac A new study used 3M person-days of Apple Watch data to develop a foundation model that predicts medical conditions with impressive accuracy.

❤️ The most beautiful art is created at the research frontier.

This new AI foundation model (JETS) detects high blood pressure (87% accuracy), atrial flutter (70%), and more.

Save hundreds of millions of lives, please. ❤️

9to5mac.com/2025/12/09/r...

Knowing what to ignore can be super helpful - for input tokens and everyone involved.

This is an important issue. It is a problem of psychological manipulation. An emergent evolution is driving foundation models toward fitness driven by usage & RLHF. Targeting dopamine instead of human outcomes, models are trained to attract users like moths to a flame, not serving and teaching.

How will humans adapt to AI?

If intelligence and labor become commodities, what will be the new human value system?

What will humans learn when they will never be as smart as their phone? What will humans do in an age of abundance?

Humans seem to be so divided in how they use AI. For some, it is like a learning and productivity companion leading to huge economic gains. For others it replaces thinking and learning, causing a downward spiral of ignorance and dependence. Is this a new skill required for human thriving?

The difference between what is possible in a lab and what is deployed by industry seems to be widening. Industry is slow to adopt AI, careful and conservative, and the timelines generally end up being longer than what the smartest people expect.

What are the fundamental sources of value in a world of ASI and robots of all types? I think primary/proprietary data and network effects will still lead to defensible margins. What else?

Today China has huge advantages in manufacturing. Will it will dominate the embodied AI trend? Will it manufacture and iterate on robots with a sustainable advantage? Or will manufacturing evolve so only supply chains, power, and physical space will matter?

Russia is so evil, continuing this genocide, and they have no intention of stopping. America and the rest of the civilized world must do everything we can to stop them.

Why are so many "news" articles posted with AI generated images? Literally fake news.

Scientists Discover That People Act Way Better When Batman Is Present It turns out that the mere presence of a guy dressed as Batman is enough to turn dead inside commuters into momentary altruists.

"the researchers had a woman who visibly appeared pregnant enter a busy train, and observed how often people offered to give up their seats. They then repeated this scenario with...a man dressed as Batman entered from another" door futurism.com/health-medic...

As I write my last column, the facts on climate crisis speak for themselves Since 1995, when the first Cop was held, carbon levels have increased from 360.67 parts per million to 426.68 parts now

Has the world given up on carbon pollution and climate change?

In 1995, when the first Cop was held, carbon was 361 parts per million. Today it's 427.

www.theguardian.com/news/2025/no...

Russia is terrorizing Ukraine. War crimes every day.

Ukraine and the world are justified in whatever they need to do to stop this.

I really enjoyed Jane Buchan on Wealthtrack.

She explains alternative investments so well.
Great education for investors.

youtu.be/QdGUUXswhu0?...

This is great. There is no such thing as too much electricity. 🤖 📈