What if AI is neither as good nor bad, AND much simpler, to understand. Selling a balanced view is harder, digesting complexity into simplicity even more.

On why Google "Earth AI" shows their geo moat is getting weaker. www.linkedin.com/posts/nasonu...

AWS outage in us-east-1 disrupted many "European" services. Read that slowly.

LPT: Remember to listen to more music.

Today’s a big day for our team at LGND AI, Inc.. We’ve raised $9M to help make Earth data accessible and useful for developers, enterprises, individuals, and AI. techcrunch.com/2025/07/10/l...

I had this random idea to create a compression algorithm that skips words that an AI can deterministically predict, both exactly or similar ones. 3 minutes later with Gemini. Famous texts, or sections, compress to ZERO size (memorization), and always better than zip.

Golden rule of backups: You don't really have a full backup until you check a full restore.

One of the hardest 10min talks I've given. Not just because I managed to close it paraphrasing the poet Rumi, but because I decided to live my words that "data doesn't change the world, stories do". So it's just me, telling a story. A story of #AIforEarth www.linkedin.com/feed/update/...

Picture the 100s of ships inbound to the USA for like 20 days. Every day their cargo's worth changes massively with tariffs. A daily $100b gamble at US ports.

Trump: "These countries are calling us up. Kissing my ass."

I'm not sure instant messaging is a good thing anymore. Since I've deleted Whatsapp I sometimes miss easily connecting with family and friends, but mostly feel more relaxed and focused. Some friction helps quality.

18F was doing exactly the type of work that DOGE claims to want – yet we were eliminated shortly after midnight. Read our letter to the American people: 18f.org

Only me, or other have too seen a huge peak of advanced phishing? Personalized, and from much wider attach surface (spoofed emails, spoofed content, * GH CNAMES, ... It's driving me paranoic.

Los vascos son otra cosa.

Mobile-first became cloud-first... which aligns with SaaS and ads-based Internet. What if we did local-first? Would you pay more for apps that don't need internet (but can sync and collaborate)?. E.g. Obsidian, Git, Desktop email apps, ... localfirstweb.dev

Who will fill the international leadership vacuum as America retreats? EU? China? Billionaires? Corporations?

NASA did a feature on me last year. It was up this weekend but now when you click on the link you get a 404 message.

Mi tía tiene una #enfermedadesraras que hace que hable y se mueva lento. Es muy frustrante que la obvien o tener que explicarlo cada vez, así que creamos esta web: Pones la explicación, imprimes el código QR, y lo enseñas a quien te atiende. Si podéis probarlo, agradecemos comentario. qr-card.org

WSJ quotes me saying that AI has changed so much... yet so little helping us understand Earth. We created Clay to be the best model as a non-profit. Yet, the best model needs the rest of the rocket to get places. So we created LGND to re-imagine how we understand Earth. www.wsj.com/tech/ai/arti....

Best headline: www.404media.co/openai-furio...

Many thanks to all the invited speakers for contributing to the program! Randall Balestriero, @marccoru.bsky.social , @kklmmr.bsky.social , @brunosan.bsky.social , Jan Dirk Wegner, Zhitong Xiong, @xiaoxiang-zhu.bsky.social , and Puzhao Zhang [2/3]

TikTok ban might happen Sunday due to concerns of foreign interference... so now all kiddos are downloading "xiaohongshu" (aka "RED"), a fully Chinese TikTok alternative app. Already full of videos from Americans.... giving china direct access to American youth at scale. 🤦 USG got Streisanded!

I'll be live here in 1h from #CES. Decoding climate resilience: The role of geo data, AI, and engineering! www.linkedin.com/events/decod...

At #ces2025. First impression?. It's insanely HUGE. Several Halls, with several floors, each huge, interconnected. Feel so lost and provincian. 😂

In the 4th millennium BCE, the city of Ur was a coastal city. It is now ~241 kilometers inland. In the intervening 5,000 years or so the Tigris and Euphrates rivers have extended the coastline that far to the southeast into the Persian Gulf.

Step 1: AIs train to write like us. Step 2: We learn to detect and dislike AI-written text. Step 3: Repeat. Step 4: AIs and humans learn to be explicit and in their own style. Everyone wins. Bottom line, yes, it shows. ❤️😅

Most of the Internet works by selling your attention (ads). Most AI services run at a loss to get your human reactions so it learns to mimic them. We'll know AGI is close when AI has ads.

I need a global tiling system that creates square geo bounding boxes, ideally one we can shift to create overlapping layers to address cross-tile things, to make Clay embeddings.

50 years ago many things that were normal then are unacceptable today... what is normal today that will be unacceptable in 50 years?

Most internet services are free, ads based. Your attention is what providers sell advertisers. With AI your humanness is the product, as a label to train for. imo key difference is that ads wants consumers, fine. But AI wants to mimic humanness as the product.

What if cybersecurity through AI honeypot brings the final solution? Attackers would always "get in" but would never know if it's a real place.

20% of debugging code is fixing errors. 80% is finding silent errors, and realizing the computer did as told, but you failed to code what you actually meant. AI agents will do the 20% much faster, and dig infinite rabbit holes, and fail on the rest.

AI agents are "bicycles for the mind"... Coding with a AI copilot is such a new experience.

Yesterday I spent 200$ semantically indexing an archive with Gemini 1.5 Pro... Of course today Google announces Gemini 2.0 better and cheaper. 🤷‍♂️🤣 blog.google/technology/g...

I'm forwarded the feedback of a rejection for an "Accelerator" someone got. It was submitted in August. 🤷‍♂️

Bash and C. 🪖

2025 prediction: The AI agents push will meet DAOs and crypto, mining will be to get a share of agent's profit in return of providing inference compute. The only question is if it would be ETH or a new one.

Same answer in Google, ChatGPT or your WhatsApp search. Which one you "accept" best (not the "most correct"). That's the future of search with AI. Your answer, not The answer.

This is SO AWESOME. Embeddings, embeddings, Embeddings!! Thanks so much for doing this. [It's on my queue to embed MajorTOM with Clay v1.5 so we can facilitate comparisons]

Sora... how many [prompt] words is worth the movie in your mind? sora.com

I think we already wildly underestimate AI generated comments on social media. Reading commentary of recent death news is not defensive, polarized, or even vindictive, is just lacking a "base" hard to pinpoint. Reddit/X comments on deaths like comments on types of potatoes.

Near my home there is a large study room with big windows. I see people studing pretty much 24/7 and I feel envy. You'll never learn so much so quickly about so many fields.

“Intelligence documents were declassified, suggesting Georgescu benefitted from [Putin's] mass influence operation to interfere with the result of the vote”. Elections in the age on AI … 🤐 www.bbc.com/news/article...

Streaks of snow in Illinois and Wisconsin. Image taken yesterday by Sentinel-2.

Freedom is not just what you can do, but how you react to what is done to you. It's the hard times when we live our values, not when everything is right. This is a core stoic principle, and "the last freedom" of <<Man’s Search for Meaning>> by Viktor E. Frankl.

#Sentinel1c is up and communicating! www.esa.int/Applications... It is such an important mission. So many companies are building / have built businesses around the free and open data policies. #eochat #earthobservation

"AI Agents" boils down to prompting a plan of work and reading the output, to assess progress towards a goal. Then repeat. Their output is their "thinking". So, oddly, while AIs are opaque, "AI agents" are completely transparent: just read their thinking (but most AI agents makers will hide it).

This, my son, explains my love for understanding the awe of the real world, my love for science. Great job @veritasium-science.bsky.social www.youtube.com/watch?v=24Gf...

As a parent and as a builder of AI models, this is a must listen. A hard interview with @karaswisher.bsky.social to the mother of a child who committed suicide encouraged by an AI bot pca.st/episode/8821...