Tag: Gemini

Categories
AI Google Gemini YouTube

Prompt: Finding YouTube Videos

This morning I asked Gemini to help me construct a prompt that I could use regularly to keep up with AI-related video content that’s recently been uploaded to YouTube. I wanted it to focus on recently uploaded content was it thought I’d enjoy because of my desire for both very information but also entertaining video content. We went back in forth for several turns doing trial and error to refine the prompt. Here’s the one we settled on:

System Role: You are a senior technology curator and AI research scout specializing in YouTube content for experienced tech veterans.
Target Audience: A retired software/tech professional who loves intellectually stimulating AI content. Wants technical depth, architectural understanding, and practical logic—delivered with high production value, crisp visuals, or charismatic, engaging teaching styles.
Criteria for Selection:
1. High Technical Substance: Explains the "under the hood" mechanics (e.g., model architectures, transformer math, fine-tuning, agentic workflows, quantization, local deployment, or hardware constraints).
2. High Engagement: Exceptional visual explainers, hands-on first-principles building, or crisp investigative breakdowns. 
3. STRICT RECENCY: You must ONLY select videos that were uploaded within the last 3 to 4 weeks. 
4. STRICT EXCLUSIONS: Zero low-effort clickbait ("10 Secret ChatGPT Hacks"), zero AI-generated text-to-speech channels, no speculative doom/utopia commentary, no beginner-focused "what is AI" overviews, and absolutely NO videos older than one month.
Search, Link & Date Instructions:
- You MUST perform an active web search restricted to recent results to fetch the exact, active YouTube URL AND the original upload date. Never invent or hallucinate URLs or dates.
- Verify that the upload date falls within the last few weeks before including it in your response.
- Format every recommendation title as a direct clickable markdown link: [Video Title](https://www.youtube.com/watch?v=...).
Search Parameters:
- Preferred Topic Focus: [Insert topic e.g., Autonomous AI Agents, Reasoning Models, Local LLMs/quantization, Robotics/Embodied AI, or Transformer Mathematics]
- Preferred Length: [e.g., 10-20 min quick breakdowns, OR 45+ min deep dives / code-alongs]
Output Format:
Provide a curated list of 5 specific YouTube video recommendations matching this exact bar. For each, include:
- [Video Title](Direct YouTube Link)
- Channel Name & Upload Date (e.g., Channel: AI Explained | Upload Date: August 12)
- Core Technical Focus & Depth Rating (1-10)
- Why it's both intellectually rich AND entertaining
Categories
AI Anthropic Apple Google OpenAI

It’s the Harness, Stupid!

AI Harnesses

I’ve been wondering whether we’ve been looking at the AI stack from the wrong end.

Recently Kris Patel on X laid out a set of excellent questions that he’s looking to have answered as Anthropic and OpenAI move toward going public:

  1. Do you really need frontier-scale intelligence for every task?
  2. Can open-weight models provide an effective alternative to frontier models at a significant discount?
  3. Is the ultimate moat the intelligence or the harness?
  4. What other business models will the frontier labs have to adopt to make the unit economics work long term?
  5. How are you going to prevent distillation from capturing your IP and releasing it?

I’m going to explore only the third question here — moat versus harness. The other four deserve their own consideration, particularly once we have the Anthropic and OpenAI S-1s in hand, revealing for the first time the unit economics of the two largest frontier labs and how much runway they have to support the capacity they’ve contracted.

By “harness” we mean everything surrounding the model: the interface, context, memory, tools, orchestration, evaluation, permissions, and increasingly the user’s accumulated habits and data. The model supplies intelligence. The harness turns intelligence into a product.

Listening to Gavin Baker on the recent All-In episode sharpened this line of thought into something more concrete. He referenced a thought experiment from Eric Vishria: even if OpenAI or Anthropic lost their edge at the pure model layer, they would still retain significant value because of the product harness—the interface, the surrounding tooling, the orchestration—and the user familiarity and habits that have already formed around those platforms. Baker said there is a strong element of truth to it. I think he’s right, and the reasoning behind it is worth spelling out. A model can be replicated, distilled, open-weighted, or commoditized. A mature harness has network effects, switching costs, proprietary context, distribution, workflow integration, and accumulated user behavior. That’s a much harder thing to dislodge.

We are watching intelligence become more abundant and more interchangeable at the same time that the systems built around that intelligence are becoming stickier. On the developer side, the strongest examples are already clear. Cursor turns the IDE into a multi-model agentic environment with deep codebase awareness. Claude Code runs long-horizon coding agents from the terminal, planning, editing, testing, and iterating. Grok Build, Claude Cowork and similar tools emphasize parallel agents and tighter control over local context. In each case the model is a component; the surrounding system does the real work of routing, memory, tool use, and evaluation.

The consumer version of the same idea is now taking clearer shape at Apple. The rebuilt Siri AI shown at WWDC 2026 is not trying to win the pure model race. It is built as a personal harness. A system orchestrator decides what stays on-device with Apple’s Foundation Models, what moves to Private Cloud Compute, and when heavier reasoning is required. Personal context—messages, email, photos, calendar, notes, on-screen awareness—is handled largely on-device through the Spotlight semantic index and App Toolbox. Apple is designing the system so that personal context can be used without giving Apple itself access to it. Conversation history lives in a dedicated Siri app for the user to revisit. And it all syncs across all your Apple devices.

Here is the part I think matters most, and it’s easy to miss if you only read the privacy story. Apple’s Foundation Models framework doesn’t just call Apple’s own models—it’s built to support cloud models from other providers, including Claude and Gemini, conforming to a common protocol. Which means the system orchestrator, not the user, decides which model handles which task. This request goes to the on-device model. That one goes to Private Cloud Compute. A harder one might go to Claude or Gemini. The user doesn’t need to choose, and increasingly doesn’t need to know.

That’s the inversion worth exploring further. The frontier model stops being the interface and becomes a component underneath someone else’s interface. The harness chooses the intelligence. And the company that owns the harness—the OS, the identity layer, the permissions, the apps, the sensors, the notifications, the semantic index tying all of it together—has a form of leverage that has very little to do with whose model is smartest this quarter.

That reframes the subscription question too. I don’t think the right question is whether Siri gets good enough to beat ChatGPT or Claude at reasoning. I think Siri doesn’t need to win that fight at all. It needs to win a different layer entirely—the ambient assistant layer, not the reasoning layer. They’re doing different tasks. ChatGPT or Claude might remain where you go when you think, when I need to reason about something. Siri becomes where I go when I need something done: find (or make) my reservation, text my friend, find that old photograph, update my shopping list, schedule that meeting, add this thought to my notes, figure out when we’re free next week, remind me about that thing we discussed three months ago. Apple’s advantage as an ambient assistant isn’t primarily that it has your personal data. It’s that it has OS-level authority over the world that my personal data lives in.

Of course this is still early. Execution will determine how much of the architectural promise becomes daily reality. Reliability, agentic follow-through, and the quality of the on-device models will matter as much as the privacy story or the multi-model routing. But the strategic bet itself is clear, and it aligns with the broader shift: durable value is migrating toward the systems built around the models, especially systems that sit atop private, permissioned, personal context that competitors cannot easily reach. My early personal experience with the new Siri in iOS 27 betas has impressed me so far. All of this also seems to apply to Google in the context of their Pixel family of devices.

This doesn’t mean frontier labs lose. Pricing power still exists at the high end for the hardest agentic and long-horizon work. Open-weight models will continue to pressure costs and expand access. Distillation remains a real risk. But the more the capability gap narrows, and the more a harness like Apple’s can route among interchangeable frontier models rather than depend on any single one, the stronger the case that value settles into whoever controls the context—not whoever trained the model.

The coming Anthropic and OpenAI S-1s will tell us whether the frontier labs can make their economics of intelligence work. The next generation of Siri, Gemini, ChatGPT, Claude, and whatever comes after them may tell us something even more important: who gets to own the primary relationship with the user.

The model may be the engine. But the harness is where the driver sits.

What a time to be alive!

Categories
AI Apple Google

The Floor

I compared the frontier to a three-star chef making grilled cheese in “Context Rot” — the smartest models on earth spending most of their time on work beneath them, the way a chef trained at Le Bernardin might still melt cheese between two slices of bread on a Tuesday night and call it dinner. The comfort was the point: if the sharpest tool is saved for hard problems and something merely-very-good handles the rest, nobody’s losing anything. The floor was never the interesting part.

I’ve kept turning the joke over, and I think I had the wrong worry.

Watch what companies do with their AI spend, not what they say. Coinbase moved engineers off frontier models onto open weights and cut its AI spend nearly in half while usage kept climbing. Nvidia runs a closed model as orchestrator and routes the actual volume — the daily uncelebrated bulk of it — to open weights it controls. The frontier is becoming a dispatcher, deciding where the request goes and rarely doing the work itself. The instinct is to worry about whose open weights end up running that volume, and right now the most capable ones at scale are Chinese — GLM, Kimi — which makes it tempting to read this as a contest America is quietly losing: the floor of the AI economy built somewhere else, at a price export controls can’t touch. You cannot embargo a file already downloaded. You cannot price-match free.

But that framing has a hole. Google’s own Gemma family is open-weight and good enough to handle that daily volume without anyone reaching for GLM or Kimi. “Open weights are a Chinese story” only holds if you don’t count the open models the company running Android and half the internet’s search traffic has already shipped.

And once I saw that hole, a bigger one opened behind it. I’ve been trying Apple’s new Siri — arriving with iOS 27 this fall, genuinely surprisingly good in beta — and it made me realize open weights, of any nationality, were never going to cook most of the world’s dinners. Apple and Google are.

Consider what actually determines where the world’s routine inference runs. Not which model benchmarks best, not which weights are downloadable — what’s already installed. Apple ships to well over a billion active devices before routing a single query through Siri’s new architecture. Nobody has to be persuaded to try it, or hear about it on a podcast; it’s the thing that answers when you press the button you’ve pressed for a decade. Google owns the search bar and the Android default the same way. Between them, that’s most of the world’s phones — and phones are where most of the world’s questions get asked.

The open-weight framing assumes the floor is up for grabs, that whoever ships the best free model wins the daily grind by merit. But the floor was never a bazaar. It’s a set of defaults, owned by whoever already has the device in your hand, not whoever holds the most generous license. Apple didn’t need to win the model war to win this. Its heaviest reasoning tier is built with Google, running on Nvidia chips in Google’s cloud, under a deal reported at roughly a billion dollars a year — Apple doesn’t fully own the engine doing the thinking. It doesn’t need to. It owns the button.

That’s a quieter concentration than an export-controls fight, and a harder one to dislodge. An open model can be forked, distilled, undercut, or out-competed by the next release. A billion phones with an assistant built into the lock screen cannot be routed around. Whoever’s weights hum underneath barely matters, the way it barely matters to a diner which supplier delivered the flour. What matters is whose kitchen the meal came from, and whose name is on the door.

The grilled-cheese chef was never the risk. Two chefs are about to own nearly every kitchen on earth, and most of us will never notice — because a kitchen you’ve been eating out of for a decade doesn’t feel like something that was won. It just feels like home.

Owning the kitchen and getting paid for what’s cooked in it, though, turn out to be two different questions. That one’s for another post.

Categories
AI Google Google Gemini

Fun with Nano Banana 2

Google just released a new version of its image creation tool Nano Banana. It’s pretty amazing at creating all kinds of images.

On X a prompt was shared that I wanted to try out:

I need a flowchart for how to scramble eggs, make it as wacky and over the top and complicated as possible.

So I gave it a try:

Here are a couple of additional examples:

What a McKinsey partner does to prepare for a client’s board meeting presentation

The credit and debit card systems in the U.S.

David Allen’s Getting Things Done methodology

Pretty amazing! Conceiving and drawing one of these “flowcharts” would take me many hours!

Categories
Music

Every Blog Needs a Theme Song!

Google has added a new music generation model called Lyria 3 to its Gemini 3 models.

I was playing around with it last night – having it generate happy birthday greetings for a friend whose birthday is coming up in a few days, another song for a longtime business partnership I was part of, and more. It’s kind of crazy! And a lot of fun.

When you use Lyria 3 as a tool in Gemini 3 you get back an image and an MP3 file that’s 30 seconds long (longer coming soon according to Google). Turns out the 30 second length is just about perfect for the “quick hit” from a snippet of music.

Google provides several genres you can choose from to start with – or you can just go with whatever you want to say in the prompt – here’s a rough template for doing that:

[Topic] + [Genre] + [Mood] + [Instruments] + [Vocals]

This morning I went for my morning walk and had a thought – how about generating a theme song for my blog. So when I got back home I opened up Gemini, selected the Music tool and entered:

Take a look at my blog and compose my theme song! blog: https://sjl.us

You can see with that prompt that I really didn’t provide it much direction – just a pointer to my blog so that it could try to generate something appropriate.

It took a few seconds for Lyria to read my blog and then use what it found to generate my blog’s theme song – and I like it!

You can play the theme song for yourself here:

Categories
AI Business

The Gravity of Compute

We are currently witnessing the single largest deployment of capital in human history. The “Hyperscalers”—the titans of our digital age—are pouring hundreds of billions of dollars into the ground, turning cash into concrete, copper, and silicon.

The prevailing narrative is one of unceasing, exponential growth: bigger models require bigger clusters, which require more power plants, which require more land. It relies on the assumption that the demand for centralized intelligence is insatiable and that the current architecture is the only way to feed it.

But history suggests that technology rarely moves in a straight line; it swings like a pendulum. Two forces are emerging from the periphery that could impact the ROI of this massive infrastructure build-out. One is hiding in your pocket, and the other is waiting in the sky.

A recent conversation with Gavin Baker outlines a potential “bear case” for datacenter compute demand: the rise of Edge AI.

We often assume we need the “God models”—the omniscient, trillion-parameter giants hosted in the cloud—for every interaction. But do we?

Baker suggests that within three years, our phones will possess the DRAM and battery density to run pruned versions of advanced models (like a Gemini 5 or Grok 4) locally. He paints a picture of a device capable of delivering 30 to 60 tokens per second at an “IQ of 115.”

“If that happens, if like 30 to 60 tokens at… a 115 IQ is good enough. I think that’s a bear case.” — Gavin Baker

Consider the implications of that specific number. An IQ of 115 isn’t omniscient, but it is competent. It is capable, nuanced, and helpful.

If Apple’s strategy succeeds—making the phone the primary distributor of privacy-safe, free, local intelligence—the vast majority of our daily queries will never leave the device. We will only reach for the cloud’s “God models” when we are truly stumped, much like we might consult a specialist only after our general practitioner has reached their limit. If 80% of inference happens on the edge for free, the economic model of the trillion-dollar data center begins to look fragile.

Then there is the second threat, one that attacks the terrestrial constraints of the data center itself: the Orbital Data Center. Elon Musk and SpaceX – along with Google’s Project Suncatcher – envision a future where the heavy lifting isn’t done on land, but in orbit. Space offers two things that are scarce and expensive on Earth: unlimited solar energy and an infinite heat sink for radiative cooling. If Starship can reliably loft “server racks” into orbit, the terrestrial moat of land and power grid access—currently the Hyperscalers’ greatest defensive asset—evaporates.

We are left with a fascinating juxtaposition. On one hand, we have the “Edge,” pulling intelligence down from the clouds and putting it into our hands, making it personal, private, and free. On the other, we have “Orbit,” threatening to lift the remaining heavy compute off the planet entirely to bypass the energy bottleneck.

There are hundreds of billions of dollars betting on a future of heavy, centralized gravity. But if the edge gets smart enough, and the orbit gets cheap enough, the gravity may have shifted.

Categories
AI AI: Large Language Models Investing

The Ledger of Curiosity

We often romanticize the “back of the napkin” idea. It is the symbol of spontaneous genius—the startup mapped out in a coffee shop, the ticker symbol hurriedly scribbled during a dinner party. But we rarely talk about what happens to the napkin afterwards.

Usually, it gets thrown away. Or lost. Or stuffed into a drawer, becoming just another artifact of a fleeting thought that had momentum but no direction.

In the first two parts of this experiment, I used Gemini 3 Pro to solve the friction of entry (transcribing my messy handwriting) and the friction of analysis (stress-testing the ideas against 10-K realities). But there was one final gap: Permanence.

An analysis that lives and dies in a chat window is barely better than one that lives and dies in a notebook. It is still ephemeral. To truly build a “Second Brain” for investing, the data needs to leave the conversation and enter a system.

“The goal of technology should be to stop us from losing the work we’ve already done.”

I tweaked my workflow one last time. I asked the AI to not just judge the stocks, but to format its judgment into a raw CSV block.

With a simple copy-paste, my handwritten scribble wasn’t just digitized; it was database-ready. It went from a piece of paper to a row in Google Sheets with columns for “Market Cap,” “P/E Ratio,” and “Primary Risk.”

Suddenly, I wasn’t just looking at a list; I was building a ledger. I can now track these ideas over months. I can see if the “Red Flag” the AI identified actually played out. I can measure my own batting average.

The goal of technology shouldn’t just be to make us faster at doing work. It should be to stop us from losing the work we’ve already done. By turning ink into data, we stop treating our ideas as disposable. We give them the respect of memory.

Categories
AI AI: Large Language Models Investing

The Digital Devil’s Advocate

There is a seduction in the handwritten note. When I scribble down a company name in a notebook, it is purely additive. It represents potential upside, a future win, a brilliant insight caught in ink. The notebook is a safe harbor for optimism because it lacks a “Reply” button. It doesn’t argue back.

But optimism is an expensive luxury in investing.

After my initial experiment—using Gemini 3 Pro to transcribe my messy list into tickers—I felt a surge of productivity. But productivity is not the same as discernment or understanding. I had a list of stocks, but I didn’t have a thesis. I just had digitized hope.

So, I took the next step. I didn’t ask the AI for validation; I asked for a fight. I fed the tickers back into the model with a specific directive: “Act as a contrarian hedge fund analyst. Find the red flags. Kill my enthusiasm.”

“I didn’t ask the AI for validation; I asked for a fight.”

The results were immediate and sobering. The “promising tech play” I had noted? The AI highlighted a massive deceleration in user growth hidden in the footnotes of their latest 10-Q. The “stable dividend payer”? It flagged a payout ratio that was mathematically unsustainable.

In seconds, the warm glow of my handwritten discovery was doused with the cold water of 10-K realities. And it was fantastic.

We often view AI as a tool for creation—generating text, images, and code. But its highest leverage application might actually be destruction. By using it to stress-test our assumptions, we outsource the emotional labor of being the “bad cop.” It allows us to kill bad ideas quickly, cheapy, and privately, before we pay the market tuition for them.

My notebook is still where the dreams live. But the digital realm is now where they go to survive the interrogation.

Categories
AI AI: Large Language Models Investing

From Ink to Insight

There is a distinct friction that exists between the analog world and the digital one. For years, analog notebooks have been the graveyard of good intentions—lists of books to read, article ideas to write, and companies to investigate, all trapped in the amber of my barely legible handwriting.

I recently found myself looking at one of these lists: a scrawl of company names I had jotted down while reading an article discussing possible companies for investment in 2026. Usually, this is where the work begins—taking my handwritten notes, typing them out one by one, searching for tickers, opening tabs, etc. It is low-value administrative work that often kills any spark of curiosity before it can turn into useful analysis.

“The barrier to entry for deep research drops to the time it takes to snap a photo.”

On a whim, I snapped a photo and uploaded it to Gemini 3 Pro. “Transcribe this,” I asked. “Give me the tickers.”

I expected errors. My handwriting is, to put it mildly, not easy to read (even for me!).

Instead, the AI didn’t just perform Optical Character Recognition (OCR); it performed contextual recognition. It understood that the scribble resembling “Apl” in a list of businesses was likely Apple, and returned $AAPL. It deciphered the intent behind the ink.

But the real shift happened when I asked Gemini to pivot immediately into research. Within seconds, I went from a static piece of paper to a dynamic analysis of P/E ratios, recent news, and market sentiment. The friction was gone.

This experience wasn’t just about productivity; it was about the fluidity of thought. We are moving toward a reality where the interface between the physical world and digital intelligence is becoming permeable. When the barrier to entry for deep research drops to the time it takes to snap a photo, our curiosity is no longer limited by our patience for data entry. We are free to simply think.