When it comes to in-game avatars, cute and cozy is underutilized as an aesthetic. Realistic avatars lead to more men and cozy avatars lead to gender parity.

Pokemon Pokopia sold 2.2 million copies in the first four days despite Nintendo going back on their word on gaming key cards. Gaming key cards are physical cartridges without game files and only DRM software. Pokpopia demand is so high that people are willing to accept gaming key cards.

In January 2026, Meta laid off 10% of their Reality Labs division. Meta shut down Horizon Workrooms (a VR collaboration app for work) on February 16, 2026. Meta is expected to layoff an additional 20% of staff and I expect a large number will come from the Reality Labs division.

Meta’s Curse of Combining Things

Meta’s weakness is their tendency to combine things that people prefer to keep separate. I assume Horizon’s avatar art direction was realistic because Meta wanted to overtake Zoom and Google Meet during the covid era as a communication platform.

The problem with this art direction is that it is less popular with female users. The closest comparison to Horizon’s art direction is Fortnite in terms of multiplayer worlds. Below is Fortnite’s estimated gender breakdown.

Let’s compare this with Animal Crossing: New Horizons.

"I think there is probably an image of 'a game where you can live leisurely with cute animals’ and ‘a game for young girls...However, looking at the gender and age data of Animal Crossing users, it is clear that the ratio of males and females is half and there are many users in their 20s and 30s."

Hisashi Nogami, Head Producer of Animal Crossing: New Horizons

Realistic avatars are great when people want excitement from a lack of stimulation in reality. But cute and non realistic avatars of Animal Crossing are more comfortable for long play times and a diversion from a stressful reality.

Pokopia maintains a similar cute and simple aesthetic.

The biggest problem with VR metaverse avatars is art direction. You want an art direction where people can customize but feel comfortable. Make things too realistic and you encounter problems with an uncanny valley. Yet, this is the direction corporations go every time.

We see this with the Apple Vision Pro whose ultra realistic avatars are even worse than Meta’s early avatars and creepy. This is done with some end game of being popular for remote collaborative work. But I believe many companies are better off with a cozy and non realistic aesthetic.

Do I think people prefer lighter things on their heads? Yes! But we’re going to run into the same problems in 2030 even with light weight AR glasses if art direction for base avatars remains the same.

Over 90 million Iranians have been cut off from the internet. International phone calls have stopped working. The Iranian government is hunting down Starlink satellite internet users (the only people that can communicate outside the country).

The Iranian casualty count is between 2,000 to 12,000 people in the past four days. For comparison, Ukraine has had roughly 14,000 casualties and the number is over 150,000 in Sudan. Media coverage for casualties is inversely proportional to internet accessibility and shareable video content.

“It is the only way,” said Mahmood Amiry-Moghaddam, co-founder of Iran Human Rights, both based in Norway. He said he received footage of protests in Mashhad, Iran’s second largest city by population.

A user in Tehran, who spoke to The Wall Street Journal through a Starlink connection early Sunday, said he had uploaded protest videos taken by relatives. He then sent them to third parties abroad who posted them on social media.

Wall Street Journal

In Italy, Cloudflare faces unappealable fines because they are refusing to cut off internet access based on whims of an Italian regulatory body. The fine is roughly double the total revenue that Cloudflare makes from the entirety of Italy.

Yesterday a quasi-judicial body in Italy fined Cloudflare $17 million for failing to go along with their scheme to censor the Internet. The scheme, which even the EU has called concerning, required us within a mere 30 minutes of notification to fully censor from the Internet any sites a shadowy cabal of European media elites deemed against their interests. No judicial oversight. No due process. No appeal. No transparency. It required us to not just remove customers, but also censor our 1.1.1.1 DNS resolver meaning it risked blacking out any site on the Internet. And it required us not just to censor the content in Italy but globally.

Matthew Prince, CEO of Cloudflare

The European Commission has issued over €20 billion in antitrust and data-protection fines since 2017. Fines are based on a global revenue and targets platforms.

Conveniently, Europe has no successful platform companies. So all fines end up being directed at American companies.

Europe should not dictate internet accessibility policies onto American companies. Imagine if a relative comes to live at your house, tells you what you do, and then takes your money for not doing what they told you. Yet, that’s what’s happening.

An open internet is the biggest safety net for all people around the world. Our country is imperfect. We disagree on a wide variety of topics. But an open internet is one issue we should agree upon.

Emerging technology can be unpredictable. I’m going to write about the unpredictability of two areas: 1) Smart Glasses and 2) AI Leadership.

Unpredictable Smart Glasses

Meta initially put cameras on glasses in 2021 to one day have augmented reality glasses. Image and video content capture was the initial value proposition to justify having cameras on glasses.

The cameras for content capture were meant to serve as a foothold for augmented reality glasses in the future. This was the plan up until 2024. By the third quarter of 2025, over a third of EssilorLuxottica’s revenue growth has come from the wearables division (aka Meta AI glasses). Meta is in the process of increasing its stake from 3% to 5% and possibly more in the future.

Early AR industry folks assumed visual data would be the main attraction. This is not the case.

The most important factor is the attractiveness of the glasses themselves and the social acceptability of wearing them. The Wayfarers are vastly more popular than Meta and EssilorLuxottica’s other frames.

As CES begins, you’re going to hear a lot about various smart glasses. My opinion is that the only smart glasses that will sell in volume will be Google and Gentle Monster’s frames. I expect most of the other glasses to be in landfills a year from now.

Unpredictable AI Leadership

At the beginning of 2023, OpenAI was on track for a monopoly in AI. At the end of 2023, Altman was forced out by OpenAI’s board. OpenAI employees synchronized tweets on x to call for Altman’s return. Many of these same people have left OpenAI as 2026 begins.

At a time when tech companies are paying eye-popping sums to hire the best minds in artificial intelligence, Google’s deal to rehire Noam Shazeer has left others in the dust.

A co-author of a seminal research paper that kicked off the AI boom, Shazeer quit Google in 2021 to start his own company after the search giant refused to release a chatbot he developed. When that startup, Character.AI, began to flounder, his old employer swooped in.

Google wrote Character a check for around $2.7 billion, according to people with knowledge of the deal. The official reason for the payment was to license Character’s technology. But the deal included another component: Shazeer agreed to work for Google again.

Wall Street Journal

Google had been bleeding talent and 2025 marked its return to dominance. Nano Banana and Veo represent the best AI image and video generation to end 2025. Google is back.

Josh Woodward, who oversees the Gemini app as well as Google Labs—a proving ground of sorts for new AI applications—called the launch of Nano Banana a “success disaster.” When people around the world began generating millions, and then billions, of images, Google was hard-pressed to find enough computing power to meet the demand. The company, he said, used emergency loans of server time to get more computing capacity.

By October, Gemini had more than 650 million monthly users, up from 450 million in July.

Wall Street Journal

Google is eating into OpenAI from the consumer side.

Anthropic is devouring OpenAI from the enterprise side with a focus on coding.

Jobs

I don’t usually write about jobs. But if you are an engineer, I’d take a look at Adaption Labs. It’s led by Sara Hooker. I’ve met her and she’s very impressive. Unlike some other leaders in AI, she displays no signs of narcissism or anti-social personality disorder which makes her unique. Take a look.

A Stanford study on political bias shows that both Democrats and Republicans view LLMs as leaning left. This represents an opportunity for an LLM model to be perceived as neutral or right leaning while everyone leans the other way.

Where is the bias coming from if it isn’t intentional? It’s coming from LLM models using scrapped reddit and wikipedia data. Reddit and wikipedia lean left. Two data points that convey this below.

There is a strong Pro-Left/Anti-Right bias to the top 100 posts of Reddit with 112x more posts (99.1%) favoring the American Left Wing compared to the Right from September 12-21, 2024. Reddit Bias

This brings me to Grokipedia. Grok is behind the other players (Google, Meta, OpenAI, Anthropic) and doesn’t have as many strong AI researchers. So how can Grok stay relevant coming from the 5th position in America?

All the major LLM models are viewed as left leaning by members of both major US political parties. This includes Grok.

Given Grok’s weaker position, I think Musk’s strategy is to capture the 46% of Americans that identify with or lean toward the Republican Party into his greater ecosystem of companies.

If Musk denies scrapping opportunities from the other major model providers on x’s data, he effectively has a monopoly on conservative leaning thought for LLM training.

Google, Anthropic, and OpenAI will continue farming reddit posts for their models - keeping their models left leaning. This will further reinforce consumer perception of model bias.

The Grokipedia strategy aims to be the neutral to conservative LLM. It explains why Grokipedia shares the Grok branding instead of a new naming convention.

Groks stays relevant by claiming it is not as biased as the other models. To do so, it needs to wean off reliance on reddit scrapping while concurrently denying scrapping opportunities on x / twitter.

Editors Note: I’m going to write more about non AI things but not writing what is on mind has led to writer’s block for a few months.

Windsurf and Cursor’s divergent paths are a great case study for AI wrapper startups. Windsurf (a leading AI coding assistant a year ago) was sold for pieces. Cursor (parent company: Anysphere) has rejected acquisition offers and remains independent after a large $900 million raise at a $9 billion valuation. How did these two AI augmented developer environment startups end up in different situations? Let’s compare.

Windsurf’s Reliance on Claude

Windsurf bet on longer context windows. It relied heavily on Anthropic’s Claude models which supported longer context windows initially. For non-technical folks, a context window can be thought of as working memory that a model considers. 100,000 tokens as a context window converts to roughly 75-100 pages.

Windsurf traded increased initial “intelligence” in exchange for reduced speed, increased costs for API calls and model dependency. This put Windsurf in a tight spot when Anthropic launched a competing product.

Windsurf and Claude Code Timeline:

• 11/14/2024 Windsurf Editor Launch

• 02/13/2025 Windsurf Wave 3 Update

• 02/24/2025 Claude Code research preview

• 05/22/2025 Claude Code Full general availability launch

• 07/11/2025 Windsurf Leadership joins Google

• 07/14/2025 Cognition acquires remaining Windsurf team

In hindsight, Windsurf was a zombie company as of February 24th, 2025.

Anysphere’s (aka Cursor) Path of Survival

Cursor’s product design routes concise, task-specific contexts to multiple AI models, rather than relying on long-context, single-model pipelines. This includes its internal tab completion model. The end outcome of these decisions is faster speed, reduced costs and reduced model dependency.

Anthropic cutting access off to Cursor would lead to a transition of API calls routed to OpenAI and Google’s Gemini models. Any decision from the leading labs to cut off Cursor’s access strengthens the balance sheet of their larger competitors. It’s not a moat in the traditional sense. But there is a form of leverage they maintain that Windsurf did not have in the short-term.

AI software development has product market fit and Cursor faces a dangerous game. Anysphere recently raised $900 million at a $9.9 billion valuation. How much of that do they use for internal model development? They can increase their burn rate and reduce future dependencies or maintain a lower burn rate while trying to increase market share in the current paradigm. Are there impending catalysts that would affect decision making?

What would you do if you were the CEO of Anysphere?

Google has acquired a 4% stake in Gentle Monster for $100 million. This follows Meta's 5% stake in EssilorLuxottica (Ray Ban, Oakley) for roughly $5 billion last year.

Why are tech conglomerates so keen on economic entanglement with eyewear companies? Eyewear is the beach head for two high priority AI strategic priorities:

1. Conversational AI usage

2. Video Data Ingestion

Apple had a stranglehold on audio with their AirPods, but glasses offer Meta a path to directly compete on audio with Apple.

Glasses are the perfect device for collecting multi-modal data through conversational AI and video / images from the real world.

Meta vs Google Alliance

Gentle Monster is a less well known brand but it has more social cachet overall and is considered a premium brand relative to Ray-Ban and Oakley. Let’s take a look at search interest for team Google (Gentle Monster, Warby Parker) vs Meta (Oakley, Ray-Ban) in Korea, Japan and the UK.

Google is positioned to be competitive in Asia with Gentle Monster alone. Although Meta is more strongly positioned for an early advantage in greater Europe. Most European countries show similar trends to the UK in terms of search interest.

It behooves Google to put a ring on it with Warby Parker metaphorically with a stake in Warby Parker. Warby Parker is an Android XR partner but there is no economic entanglement.

China based Huawei collaborated with Gentle Monster on Smart Glasses in 2020 and 2021. They now find themselves on the outside looking in. The biggest beneficiary of Huawei getting squeezed out is likely going to be Qualcomm.

Google should act fast to avoid getting squeezed out of Warby Parker. The same way Huawei was squeezed out with Gentle Monster despite an early collaboration.

Samsung’s Mimetic Errors

Samsung’s big weakness is they only act after Apple acts. My take is that because Apple hasn’t acted yet on glasses, Samsung hasn’t either. A mimetic error.

Samsung’s HMD device (Project Moohan) looks better and is more reasonably priced than Apple’s Vision Pro. The problem is the AVP (and HMDs) didn’t take off. This is going to end up being an unused Christmas gift.

Founder Led Companies

I saw a tweet from Talia Goldberg on the difference in performance between founder led companies vs non founder led companies. The graph blow is striking.

Meta is in the lead. It’s also the only company with an active founder CEO.

Google has founders that poke around but are not active in running the company. Google is second in the AI glasses race.

Apple is lagging behind both. There are people out there that will say but Apple never comes out with product first and still wins the market. To which I say, this was true with Jobs. Under Cook, the ability to enter late and win has not continued.

Samsung is likewise a step behind because they are imitating the wrong competitor.

What about Snap?

Snap is still founder led. They all-in on AR glasses. Snap does not have eyewear brand partnerships and has a consumer focus.

The reader base of this newsletter is more likely to be interested in AR than the average person. Yet only 43% of readers indicated that they would wear the Meta Orions daily.

People have very high standards for what they are willing to wear on their face daily. I won’t count out Snap because they are still founder led.

Eyewear brands represent real estate on people’s faces. This future real estate is the key to conversational AI dominance and real world data ingestion.

Apple and Snap do not have eyewear brand partnerships yet. Do you think they are necessary? Or are they strong enough as consumer brands to compete solo?

One looming event on my mind is something I’ve been calling the AI Event Horizon. AI growth is limited by people. This limitation is removed when we have automated AI research powered by nuclear power plants. Just as light cannot escape a black hole, human information cannot escape AI the closer we approach the AI Event Horizon.

The AI event horizon is the point in time when two conditions are met:

• Automated AI Research: AI performs the work of AI researchers and can recursively improve itself without humans

• There is nuclear energy to power the automated AI researchers at scale

This is going to be one of the most significant events in human history. Automated AI research requires AI researchers. This explains the explosion in compensation packages for AI researchers.

CEO Mark Zuckerberg himself has taken to emailing job offers to elite talent. One AI researcher told us Zuckerberg offered them an eight-figure compensation package – at least $10,000,000 a year.

"I got an email from Mark personally," our source said. "And he said, 'I have an offer for you.' Wow, and the offer was crazy." The Register

Once the human AI researcher rate limiter goes away, the next bottleneck becomes nuclear energy.

I first shared this chart below last year in my Securing American AI Superiority article. Since then, Constellation Energy’s stock price has gone from $216.47 to $309.02. NuScale Power has gone from $12.53 to $44.07 (see the first graph above).

The growth in stock price for Constellation and NuScale is a product of anticipated energy demand for AI data centers exclusively. No one other energy source can power the energy demands that are about to come for AI.

After I wrote Securing American AI Superiority last July, two deals involving technology companies announced deals with Constellation Energy.

Microsoft

After 17 months, Constellation and Microsoft announced in September that the tech giant has signed a 20-year power-purchase agreement to pair its data centers with round-the-clock clean power from Three Mile Island. Constellation expects to spend around $1.6 billion on the unprecedented effort to resuscitate a reactor already undergoing decommissioning. Wall Street Journal

Meta

Today, building on our efforts to support next-generation and advanced energy technologies, including geothermal and nuclear, we are announcing a 20-year corporate nuclear energy agreement with Constellation Energy for the Clinton Clean Energy Center. Meta

Constellation represents actual demand for nuclear energy while NuScale is more speculative.

This surge in demand in nuclear energy companies is indirectly a response to another chart. This one is more speculative.

If the timeline above is accurate, we will see the nuclear energy companies continue to rise. If the nuclear energy stocks flatline, it implies that the timeline for the chart above might be extended into the future a little longer.

As we get closer to the AI Event Horizon, we’re going to undergo some big changes in human society. I can’t fully articulate exactly what will happen.

Here are a few thoughts rotating around my mind:

• Universal Basic Income (first proposed by Andrew Yang) - will be taken more seriously by people across the political spectrum once we cross the AI Event Horizon

  • Work in some sense will become more a choice in terms of finding purpose and requiring a greater sense agency

• The new standard for white collar work will become greater than AI

  • This means better than AI at one thing

  • This might also mean requiring human trust

  • Or it could also refer to human connection and relationships

• There’s a solid two to three years before the world gets really weird, go after your priorities now and push hard with focus and intentionality

What do you think will happen in America after the AI Event Horizon?

I’m looking to incorporate well thought-out responses from readers in the form of quotes and possibly video chat interviews (if we can scheduling to work out).

One of my favorite things about paper magazines in the old days is they would includes letters to the editor (which were typically a few sentences). I’ve been brainstorming on how to do something similar but more contemporary.

Scopely has acquired all of Niantic’s major gaming properties. Scopely was acquired by Savvy (the gaming unit of the Saudi Arabian Public Investment Fund) last year and effectively has an unlimited balance sheet.

First, for Niantic Inc., we have reached an agreement for Scopely, a world leader in mobile games, to acquire Pokémon GO, Pikmin Bloom, Monster Hunter Now, and the incredible teams working on these experiences, for $3.5 billion from Scopely with an additional $350 million of cash from Niantic being distributed as part of the deal, yielding a total value of approximately $3.85 billion for Niantic equity holders.

Niantic

The new Niantic Spatial focused team comes out with a clean $250 million balance sheet. What remains is a geospatial mapping team that:

• Moves out of the 1st party gaming development business (consumer)

• Refocuses on providing geospatial mapping services to robotics, autonomous vehicles and AI teams (enterprise)

Coatue previously invested $300 million at a $9 billion valuation. Niantic comes out with investors that are likely all made whole and playing with house money now. It’s unknown what the liquidation preferences look like for different parties on the cap table.

Between 2015 and 2021, Niantic secured at least $770 million in known equity investments, with Coatue, IVP and Spark Capital among its lead investors.

Nasdaq

Future financing comes without needing to address the volatile nature of the gaming business (ie Wizards Unite). Niantic NewCo becomes an AWS for geospatial mapping and AR development.

Niantic will retain the Ingress and Peridot properties that were internally developed IP. These properties will serve as consumer facing avenues for demo’ing new features developed by the remaining Niantic R&D team.

Some implications:

• Pokemon GO

  • Remote raiding increases - Scopely has no mission or incentives towards pushing for in-person raiding.

    • Most PokemonGO players probably like this.

  • Monetization increases - Are unskippable ads coming soon to PoGO and Monster Hunter?

    • Most PokemonGO players probably don’t like this

Do any of you still play PokemonGO? If so, what do you think of the trade-off between more remote raiding and more ads simultaneously?

Experiments

This year, I’ll be experimenting more with content at the end of essays and articles. Some of this content will come and go.

IRL Note

I’ll be at the a16z speedrun demo day next week if anyone is attending and wants to catch up in person.

Easter Eggs - Music

I will periodically share music I listen to while working.

Let me know if you like any songs!

Share

Subscribe now

Volodymyr Zelenskyy, Justin Trudeau, and C.C. Wei. Two of these people are narcissists and only one of them is an effective leader.

Three examples and three outcomes with interactions with Trump:

• Zelenskyy: Negotiations over Ukraine’s non existent rare earth metals

• Trudeau: Canadian Tariffs

• Wei: Taiwan Semiconductor Manufacturing Invest $100 billion into U.S.

Ukrainian Imaginary Rare Earth Metals

The most bizarre news story right now is the ongoing Ukraine rare earth metals coverage. Ukraine doesn’t have any any valuable rare earth metal mines. 50% of zero is still zero.

Donald Trump is foolish for asking for something that has no value.

Volodymyr Zelenskyy is an idiot for not immediately agreeing to give away something worthless in exchange for hundreds of billions of dollars.

These rare earth metals are valuable.

• Neodymium and Dysprosium: Needed for high-performance magnets used in electric motors and cooling systems in data centers.

• Praseodymium: Needed for magnetic alloys in hardware.

• Europium and Terbium: Needed for displays and lighting of advanced computing systems.

Ukraine has less than 1% of important rare earth metals. Their rare earth metal mines are scarce and unprofitable to mine.

What Ukraine has is grain, iron and coal. Instead of rare earth metals, negotiations should be focused on a percentage of Ukraine’s grain, iron and coal. These are resources Ukraine has and they have value.

On the other hand, Ukraine’s rare earth metals don’t exist or are unprofitable to mine - making them worthless. The coverage by the media and the negotiations between Trump and Zelenskky on this topic are completely asinine.

Let’s assume rare earth metals are important. Let’s also assume China is America’s greatest threat. The most strategic thing to do is to move US military resources to Asia and strengthen ties with Australia.

Neither Ukraine or Russia is relevant. Zelenskyy overplayed his hand with his arrogance.

Canadian Tariffs - Bad for Canada, Good for the Liberal Party of Canada

The tariffs are the 51st state trolling are a product Trump’s hate of Trudeau. That’s it.

The brief video posted on Twitter by Canada's public broadcaster, CBC, showed Mr Trudeau chatting with a group of leaders, including Mr Johnson, Mr Macron, Dutch Prime Minister Mark Rutte, and Queen Elizabeth's daughter Princess Anne at Buckingham Palace.

At the start of the footage, Mr Johnson asked Mr Macron: "Is that why you were late?"

Mr Trudeau then interjects: "He was late because he takes a 40-minute press conference off the top."

BBC December 4th, 2019

Trump called Trudeau two-faced afterwards and has hated him ever since. Trump is spiteful and holds grudges.

The quickest solution to avoid this problem would have been Trudeau resigning last year.

Trump’s tariffs are ironically beneficial to polling for Canada’s Liberal party because it galvanizes patriotism around the current leader and party.

There is a prisoner’s dilemma in play here. The best way for the Liberal party to win the next election is for the tariffs to continue. The tariffs are most likely to continue if Trudeau stays in office longer.

The dance that is currently in play is one where Trudeau stays in office just long enough for the Liberal party to overtake the Conservative party in Canada in polling without completely ruining Canada’s economy by the time it happens.

A real leader would have done what is best for his country instead of his party.

TSMC Invest $100 billion into U.S.

Taiwan Semiconductor Manufacturing will pour $100 billion into bolstering chip manufacturing in the U.S., President Donald Trump announced Monday.

The president called the investment a “tremendous move by the most powerful company in the world.” The new capital brings TSMC’s total investment in the U.S. to $165 billion and will go toward building five new fabrication facilities in Arizona.

CNBC

This is a really smart move by TSMC. As I mentioned last year, almost all advanced chip manufacturing occurs in Taiwan.

The image below is from my Securing American AI Superiority article last year.

This $100 billion investment serves two purposes. It prevents tariffs on Taiwan. More importantly, it is an indirect protection fee for the entire country of Taiwan.

China’s military has set up a zone for “live-fire training” about 46 miles (74 kilometers) off the southwestern coast of Taiwan without advance notice, the island’s defense ministry said on Wednesday, condemning the move as provocative and a threat to international navigation. It comes a day after Taiwan’s coast guard detained a Chinese-crewed cargo shipsuspected of cutting an undersea cable in the Taiwan Strait.

CNN

If you are Taiwanese, give thanks to C.C. Wei, CEO of Taiwan Semiconductor Manufacturing Company. He saved Taiwan from tariffs and probably assured U.S. military protection for the near future at a time when China is increasingly aggressive.

This is what a real leader looks like. One narrative that I disagree with is that CEOs have too much power. CEOs are responsible for doing things that work. Politicians just win popularity contests.

Volodymyr Zelenskyy, Justin Trudeau, and C.C. Wei. You don’t necessarily want a politician responsible for the fate of your country.

Share

Subscribe now

My estimate is around 75% of the folks that I knew in augmented reality (AR) have transitioned to artificial intelligence (AI) organizations or divisions in the past two years. Why is this happening?

Let’s go through an example of trying on clothing with AR. An AR shopping experience used to involve creation of 3D assets by a creative or technical artist. The 3D asset had logic for repositioning in real-time as the user moved around. A developer was responsible for this second portion. AR clothing try-on and AR filters led to increased product purchases.

It was effective and interactive marketing. A year and a half ago, Meta announced a new AR ad format. A year ago, the AR ads were cancelled. Why would something that worked well and required years of community building be cancelled?

Let’s compare AR clothing try-on with AI clothing try-on. A static image of the person is uploaded. Algorithms are then used to blend the clothing item onto the image with the user. This process uses less computing resources (50-70%). It also doesn’t require anyone to build a 3D asset or rig it. Rigging is the creation of a skeleton for the 3D asset.

Generative AI clothing try-on costs less to make, looks better and requires less computational resources relative to traditional AR.

Fashion and entertainment industry marketers are two of the most technology forward groups. Many AR studios and independent creators were indirectly sustained through AR marketing experiences serving the fashion and entertainment industry. That lifeline is gone. Along with it, nascent AR gaming studios have lost a core source of revenue through consulting projects.

What happened to the talent that was in AR?

First the product professionals (engineers and product managers) moved to AI because there were more resources and jobs. This started roughly two years ago.

The AR distribution professionals (marketing, community and developer relations professionals) moved to AI after the new AI products needed users. This AR to AI distribution professional migration started last year.

Finally the distribution professionals reached out to early adopters of AR to use the new tools. This is happening as we speak. Below is an early AR adopter that has since been pulled to create demos for new AI tooling.

Here’s my guess on what will happen this year. Generative AI for creative use cases will proliferate. The different generative AI teams will ramp up resources for their distribution teams.

These distribution teams will ramp out reaching out to early technology adopters. A large portion of these early adopters will be former AR creators. This is because the distribution teams in AR divisions have transitioned to work in AI. The first people they will reach out to are people they used to know in AR.

The AR ecosystem top to bottom has been swallowed whole by generative AI. I’m still interested in AR. But if you are wondering I am writing less about AR, now you have your answer. There’s simply less to write about because there are less people working on AR.

The AR industry has been cannibalized by the AI wave for the time being. It’s not the first industry impacted and it won’t be the last.

Share with anyone that you believe would enjoy reading this article.

Share

If you are not subscribed to Games of Strategy, subscribe below.

Subscribe now

Descriptions of Images

1. Image One: Flair AI

2. Image Two: Lens Studio Garment Texturing

3. Image Three: Josephine Miller using Meta’s Segment Anything 2 Model

The image above is from the Apple App Store. An open source AI research team called DeepSeek has released competitive and popular new foundational AI model. There are parts of the DeepSeek story that are questionable and other parts that are undeniable. I go through all of this.

Undeniable

One undeniable part of the story is that open source AI models are here to stay. There was a period when some people believed closed sourced models backed by greater resources had an unreachable advantage. It seems like open source AI is here to stay. For all the flack Mark Zuckerberg gets, Meta’s Llama is a big reason open source AI has been able to stay relevant.

Another undeniable part of the story is that DeepSeek’s model is very cost effective. Large language models (LLMs) are priced on a token basis. A token is a small portion of text that is typically a couple of letters.

DeepSeek R1

• $0.55 per million tokens (input)

• $2.19 per million tokens (output)

OpenAI o1

• $15.00 per million tokens (input)

• $60.00 per million tokens (output)

DeepSeek provides a compelling product that people are using. Otherwise, it wouldn’t be number one on the App Store.

Questionable

The questionable part is whether roughly $6 million dollars and 2,048 NVIDIA H800 chips were used to train the model. It’s possible but seems unlikely in my opinion.

The number seems small. The DeepSeek team is likely able to avoid advanced chip export restrictions imposed by the outgoing Biden administration but can’t talk about it publicly for obvious reasons.

By the close of 2024, the Bank of China is set to provide loans amounting to 1.91 trillion yuan (around £195 billion) to more than 100,000 technology companies.

Bank of China Innovation Plan

The Bank of China would not be handing out so much money to Chinese technology companies if training foundation models was so affordable. This is just me using simple deductive reasoning.

Implications

One important implication is throwing money at AI alone doesn’t ensure winning. The Stargate project is a $500 billion AI infrastructure project recently announced.

This large sum of money does not ensure victory. The other big implication of the story is that AI is still a wide open game.

It’s in America’s best interest to bring the best minds of other countries to America and keep them. Keeping talent out of the country just creates competitors outside of the scope of the American hegemony. It’s a lot more wise to brain drain other countries of their top minds as we enter a critical phase of AI reshaping the world.

Subscribe now

Share

One interesting observation through the TikTok shutdown saga is the value of Twitter (aka x). Twitter stands despite legacy’s media best efforts to kill it post-Elon acquisition. The reason why mainstream media outlets can’t kill twitter is because it is still the fastest source of breaking information of all the social platforms.

On Instagram people were mourning the loss of TikTok heading into this weekend. Twitter users engaging on this topic were aware that Shou Zi Chew (CEO of TikTok) was headed to the presidential inauguration. They were also aware of Trump’s pronouncements to save TikTok.

There was between a half day to a two days of information asymmetry on TikTok’s ban coverage for twitter users relative to Instagram. Breaking news is partially more prevalent because of anonymous accounts. Anonymous accounts are a double edge sword because it also comes with increased toxic behavior and trolls. Twitter will fundamentally always be a more toxic place if it wants to retain it’s breaking news identity.

Instagram Edits - A rare situation where Meta was too slow to imitate a competitor

Meta recently revealed Instagram Edits. Edits is a video editor meant to compete with TikTok / Bytedance’s CapCut video editor.

CapCut was available internationally in 2020. It had 200 million active users in 2022. CapCut has a billion downloads from the Google Play Store alone as of January 2025.

Meta / Instagram Leadership’s two biggest strategic mistakes against TikTok were:

1. Not focusing on making the best video editing tool

2. Promoting TikTok indirectly by showing videos with TikTok watermarks in a desperate initial catch-up effort

The platform with the best content creation tools gets the best content first. When you make it easy for people to edit quality videos on your app, they use it to record and edit more. They share the edited content first on the platform they first do the editing on.

TikTok’s superior video editing meant that social video creators have started their user journeys on TikTok first for the past four years. This also led to the TikTok watermark proliferating throughout Instagram. The strategic mistake of the product team of Instagram was not understanding this basic user flow from the perspective of the content creators.

Almost every early Reels video on Instagram had the TikTok watermark because Instagram did not develop better video editing tools earlier. If Instagram created Edits earlier or focused on Instagram’s default video editing tools, TikTok would not have been able to monopolize the market of short form video content creators.

Instead Instagram was effectively paying video content creators to share videos that were first made on TikTok. They were paying creators to promote TikTok because all the early Reels videos had a TikTok watermark for almost two years.

One thing Edits has going for it is video analytics integrations. Analytics is what has allowed Youtube to maintain long-form video supremacy. Instagram Edits’ preview page on the Apple App Store suggests that content creators will be able to see some form of analytics associated with Instagram video performance if they use Edits.

The better the analytics, the more video creators can optimize their videos. This in turn will subconsciously make content creators focus on optimizing their Instagram videos. Especially if Instagram shares data on the video that is not accessible anywhere else.

Edits might lure away some video creators since CapCut moved from being a free tool to one with a subscription fee. This free offering might be able to draw content creators just starting off.

Virality and view counts is lower but Instagram still commands higher compensation for content creators.

Can Edits catch up? It will need to if Instagram wants to catch up in video. Otherwise Instagram’s recent move to turn all Instagram square grids into rectangles will be for naught.

Do you use CapCut? If so, I’d love to learn what it would take for you to move to competing free platform that was free.

Subscribe now

Share Games of Strategy with Kenneth To

Modern consumer augmented reality was popularized by face filters on social media.
Face filters allowed everyday women to share more content. Face filters reduced the friction to content creation:

• Reduces time spent on makeup touchups

• Reduced need for location switching for content creation

Tomorrow Meta shuts down their face filter creation platform which led to increased app usage. Why would Meta abandon Spark AR? Just last year they were going to integrate AR filter ads into their ad system. Why cancel it? The answer is both questions is AI.

Meta wants people to use their generative AI features as a priority. Here’s what I think happened. Face filters and consumer AR usage on Instagram and Facebook became a distraction from the new goal of increasing generative AI usage. That’s it.

Meta’s first consumer AR ecosystem became casualty of changing tides at a macro level.

As 2025 begins, I see Meta doubling down on AI. On the hardware side, this means glasses. Over the past year, Meta has been rebranding smart glasses to AI glasses. This is an important rebranding considering that Smart Glasses were anticipated to be a precursor to AR glasses. This is connected to the Spark AR shut down indirectly. Meta’s priority is to get people using their AI models. This priority influences all strategic decisions across all teams and all verticals.

The Rayban Meta Glasses have performed better than expected. I think this is the angle Meta sees for supplanting Apple. Not through the Quest but through the Glasses. Sometime over the next decade, we will see AR re-integrated into the smart glasses. But we are about to enter a dry period.

The Rayban Meta Glasses will succeed in a way that a lot of the competitors will fail. My video with Kiara alludes to the reason but women care about aesthetics and branding more than men on average.

Quick thoughts on CES. Last year at CES we saw a lot about AI pins and pendants. These products all failed. You are going to see a lot of articles about smart glasses. I’m sharing an opinionated take that other reporters won’t share.

• Vuzix is a zombie company and will be sold off for IP eventually

• XREAL will bleed money and not take off

I’m saying this as someone who is a big fan of AR and big believer in AR glasses in the long-term. Both Vuzix and XREAL lack fundamental branding and aesthetics to take off. Meta has a partnership with EssilorLuxottica - the top eyewear conglomerate in the world.

On a relative basis, opposing smart glasses companies will face a market where they have inferior branding and software. Is it possible to overcome this? It’s not zero but it’s not very likely.

Part of a tech reporter or influencer’s job is to keep this “anything can happen” perspective in order to maintain access to sources. Or maybe the journalists and influencers are not aware of the odds at all. But the portrayal of anything can happen with smart glasses I am seeing from CES is off. I think I know what will happen. It’s going to come down to Meta, Snap and Apple with Google on the outside lane looking in.

A note on this newsletter and what to expect in 2025.

I’m going to write more frequently across the domains of emerging technology, media and culture. Last year I didn’t write consistently because I was worried about people not wanting to read about AI. But a lot of these technology topics are interconnected because the companies that are driving the technologies are conglomerates.

If this new direction is not for you and you decide to unsubscribe, I wanted to express my gratitude to you for taking the time to read my essays the past few years. For those of you that remain, expect a higher publishing consistency but across a broader spectrum of domains. What will remain consistent is that I am committed long-form writing and strategic analysis.

I’ll also try to interview more people and introduce more perspectives from experts you don’t get from mainstream media. Traditional media sources interview the same people and get the same bland PR friendly non answers as answers. I want to share non-consensus perspectives.

Subscribe now

Share

Meta’s Orion AR glasses signal the beginning of the AR glasses era. Software has been ahead of hardware in augmented reality (AR). Hardware is what has been holding back adoption of AR.

Why hasn’t AR taken off yet? Let’s go through hardware options for AR one by one to understand what has been holding back AR up until now. Then we can contemplate why AR glasses can replace smartphones as people’s primary personal computing device.

Smartphones

The most popular use case for AR for the past six years is face filters on social apps. All the major players in the social space (Meta, Snap, TikTok) have tried to push worldspace AR. The usage rates for back facing worldspace AR experiences pales in comparison to the front facing camera AR filters.

Smartphone aren’t designed for outdoor AR usage. Any extended outdoor AR usage leads to overheating and auto-brightness dimming. AR on smartphones is still a gimmick because of overheating and auto dimming which prevents extended AR usage. The most popular use case of AR being selfies and short form video content recorded indoors.

Head Mounted Displays (HMDs)

HMDs are closer to the ideal form factor for AR. The Quest isn’t supposed to be used outdoors. Neither is the Apple Vision Pro. Indoor AR usage is suboptimal because of space and product distribution via sharing of content.

Most people are at 10x10 or less in free space. Indoor AR is effectively an inferior visual experience compared to VR with none of the benefits.

Indoor space is intimately tied with product distribution. People with smaller spaces are likely to be more self conscious about sharing gaming footage from inside their homes. So you end up with a paradigm where no one shares HMD AR content online.

Developers of consumer AR experiences can’t get funding and they also can’t get viral product distribution that would lead to funding.

AR needs attractive glasses designed for outdoors

Meta’s Smart Glasses have proven people are willing to wear attractive smartglasses that have cameras on them for extended periods of time in public.

What we’ve learned from Meta’s Smart Glasses is people are willing to wear computers in the form of glasses. They just need to look like regular glasses as much as possible.

The Orions don’t look like the Meta Smart Glasses yet. But they are very close.

The Apple Vision Pro weighs between 600 to 650 grams. The Quest 3 weighs 515 grams. Meta’s Orion smartglasses weigh 100 grams.

At 100 grams, you have something suitable for all-day wear.

Estimates are that the glasses cost between $10,000 to $20,000 to produce. So it’s currently too expensive for mass consumer. Production costs need to go down and they will over time.

Most importantly feasibility is solved. For a long time, AR developers have talked about a future where attractive AR glasses replace smartphones. But the actual glasses did not exist. The future has arrived.

What happens next?

Here’s what I predict happens next:

• An even more aggressive onslaught by Apple on digital advertising to reduce Meta’s and Snap’s revenue

• 3rd Faction will form around Snap

  • Most 3rd party developers are not incentivized to have Apple win AR glasses war

    • Apple’s App Store monopoly and 30% app tax on smartphones would continue

  • Snap being the smallest of the the major AR glasses players is a more acceptable ally in terms of relative leverage

If you know someone interested in AR, I’d appreciate it if you shared this piece with them!

Share

Have any questions or thoughts about AR glasses? Let me know.

Both the United States and the European Union have rolled out FLOP restrictions. FLOP is shorthand for floating-point operation.

What exactly is a FLOP?

A FLOP is a simple math calculation that computers solve. Imagine a book of math problems. Every problem involves basic arithmetic. Every time a computer solves a problem - it does a FLOP.

The problem with FLOP restrictions is that it likely to fail to mitigate risk at all. See the following excerpt from Sara Hooker’s latest paper on compute thresholds with FLOPs as a proxy.

Both the White House Executive Orders on AI Safety (EO) and the EU AI Act encode the use of FLOP or “floating-point operations” as a way to identify more powerful systems. What is striking about the choice of compute thresholds to-date is that no models currently deployed in the wild fulfill the current criteria set by the EO. This implies that the emphasis is often not on auditing the risks and harms incurred by currently deployed models – but rather is based upon the belief that future levels of compute will introduce unforeseen new risks. A key conclusion of this essay is that compute thresholds as currently implemented are shortsighted and likely to fail to mitigate risk.

On the Limitations of Compute Thresholds as a Governance Strategy. by Sara Hooker

Models for different things have different relationships with scaling and FLOPs. Higher compute models with FLOPs as a proxy generally scale better with language models. This suggests to me that any single number for a FLOP restriction is completely arbitrary at best and doesn’t serve the claimed premise of safety guard rails at all.

Established AI companies with more resources can comply more easily with FLOP restrictions - that they directly dictate through lobbying resources. This creates barriers for smaller or new entrants who may struggle with the costs and complexities of meeting these regulations. FLOP restrictions become a form of regulatory capture by larger first movers. This is anti-competitive regulation under the guise of providing safety.

Unlike America and Europe, China is working to actively increase compute.

China plans to increase its computing power by 50% by 2025, the country’s key ministries said Monday, as it looks to keep pace with the U.S. in artificial intelligence and supercomputing applications.

The world’s second-largest economy wants to have computing capacity equal to 300 exaflops, according to a plan from six government departments, including the powerful cyberspace regulator. That would be up from the 197 exaflop computing power the country currently has.

“China has found that traditionally, every 1 yuan invested in computing power has driven 3-4 yuan of economic output,” Akshara Bassi, senior research analyst at Counterpoint, told CNBC via email. “The investments echo China’s plans to drive economic output through leadership in technology prowess and integrating AI with existing technologies and solutions across all industries and domains.”

CNBC

This is a negative trend for Americans. We must undo these FLOP restrictions in my opinion.

If you enjoyed agree with me, please share with someone you know. It’s important that America does not fall behind in AI with stupid policies.

Share Games of Strategy with Kenneth To

We need to undo this executive order on FLOP restrictions from last year’s executive order. I’m fairly confident that Biden had no idea what a FLOP is when he signed this. We can’t allow lobbyists to enact regulatory capture and sabotage American leadership in AI.

Have any thoughts or questions on AI or FLOP restrictions? Let me know. Also open to perspectives I may not have not considered.

AI super intelligence might come a lot faster than anticipated. AI advancement requires four resources: 1) information 2) AI researchers 3) energy 4) compute. Here’s my brief thoughts on each. I’d love suggestions for papers or books to read if anyone has any!

Information

I’ll touch on information first because I think a big portion of new information and data will come from smart glasses and direct messaging platforms. Meta has both. Tech reporters skeptical of Meta’s investments in Reality Labs show their ignorance because they don’t understand the importance of multi-modal data ingestion for AI training superiority.

The internet has already been scraped by every player. Marginal utility is reaching an asymptote from internet scraping. You need superior proprietary data.

Unique sources of information will be the key advantage for different players. Microsoft will obtain this through Windows and their office suite. Google will obtain this through google’s suite. Apple through the iPhone.

AI Researchers

AI researchers themselves can be modeled and automated. We can train models to do all types of things but the biggest thing is modeling a researcher. This is my biggest insight from reading Situational Awareness over the weekend.

AI researcher models will be able to work day and night. Together. In adversarial models against each other. The limitation will be energy to run the compute and the chips for the compute. I already touched on information, so let’s go to the other two.

Energy

The first roadblock will be energy. America needs to create more energy to sustain running the AI farms. That’s why nuclear energy companies like Constellation are surging.

Energy needs to be scaled up. The only sustainable solution is nuclear energy growth.

Compute

Let’s take a look at compute. In particular, let’s see where advanced AI chips are being made.

Compute represents the largest risk factor. It’s all in Taiwan which faces a threat of an invasion. American companies need to de-risk this.

They need to push advanced chip manufacturing to move to North America where it can be protected by America’s nuclear arsenal. Until then, Taiwan is Dune with advanced chips representing Spice in this analogy.

Advanced chips for AI are a critical resource upon which everything else requires. Right now, American tech leaders have no strategy other than praying no invasion happens. They need to be more proactive here.

I’m interested in strong educated opinions on the topics of 1) information 2) AI researchers 3) energy 4) compute for building models. Would love to chat if you have any thoughts!

The graph above shows search interest for the Meta Quest vs Apple Vision Pro (AVP) for the past year. The majority of tech reporters were seemingly wrong about how Quest vs AVP interest would play out.

Here’s what I think is happening:

1. The Quest got popular in the holiday season last year with Meta selling effectively at cost

2. AVP surged because it is a) New b) Apple has the best brand C) Tech reporters favor Apple and hate Meta

3. People lost interest in the AVP because a) price and b) lack of apps

Let’s take a look a look at search interest for the most recent 90 days and include the Ray-Ban Meta Wayfarers, the Rabbit R1 and Humane’s Pin.

The past 90 days are an interesting barometer because it is far away from announcements and the holiday season.

XR hardware thoughts:

• The Quest has taken a commanding lead in XR hardware during this non-news period

• Quest early interest means that XR indie developers will mostly be developing for the Quest

• Apple will seemingly focus on enterprise use cases. They are giving raw camera access to enterprise only. Apple could have had a surge by opening this up for consumer apps but chose not to.

AI hardware thoughts:

• Meta’s Smart glasses with the Wayfarer design are day-to-day socially acceptable.

• I did a query that included the Meta’s Headliner design but interest in those are negligible compared to the Wayfarers. Design is critical for smart glasses.

• No one wants the Rabbit or Humane AI devices. Don’t buy them, these companies are dead.

• Smartglasses are a more ideal form factor for AI first hardware relative to pendants and pins.

I have two questions for readers.

Apple is holding off on a AVP 2 and diverting resources to a more affordable HMD. At which price point would you prefer it over a Quest 3 ($500).

Answering polls like this helps me communicate with executives and founders about what people actually want. Thanks for taking the time to participate!

If you have any questions that you think would make for interesting polls on emerging technology, let me know and I’ll consider it for future articles.

The Apple Vision Pro (AVP) is Apple’s first attempt at a commercial spatial computer. I think a lot of coverage on this device is off and lacks understanding of both personal computers and Apple’s current position.

Here’s my take on the Apple Vision Pro along the following topics:

• Apple Vision Pro as a part of Apple Computer History

• Apple Vision Pro and Category Definition

• Apple Vision Pro’s Lack of Apps and Developer Relations

This launch is very different from previous product launches. I’ll explain why.

Apple Vision Pro as a part of Apple Computer History

A personal computer is a stationary computer. The Mac represents Apple’s flagship product in the laptop category.

A laptop is a portable desktop computer. The MacBook represents Apple’s flagship product in the laptop category.

The form factor got smaller from a Mac to a MacBook.

A smartphone is computer that fits in your pocket. The iPhone represents Apple’s flagship product in the smartphone category.

The form factor got smaller from a MacBook to an iPhone.

A HMD (head mounted display) is a computer that is worn on your head. The Apple Vision Pro (AVP) represents Apple’s flagship product in the HMD category.

The form factor got larger from the iPhone to the Apple Vision Pro.

This form factor pattern change in size will likely affect adoption. Portability seems to be a factor in adoption of new personal computing devices. A device that reverses this trend will get different results.

Apple Hardware and Category Definition

The Mac did not redefine the personal computer category. It redefined usability by free reducing crashes and computer viruses relative to Windows based PCs.

The MacBook Pro did not redefine the laptop category. It redefined usability by improving battery life, stability and UX consistency.

The iPhone reimagined and redefined smartphones. Prior to the iPhone, smartphones came with styluses and keyboards. The iPhone stripped away all peripherals with a unified touch interface. The iPhone is also Apple’s most successful product.

The AVP aims to reimagine and redefine the XR HMD category. Prior to the AVP, XR HMDs came with controllers. The AVP strips away the controllers.

The Apple Vision Pro is not a direct corollary to the iPhone though. While it takes away controllers, it also adds a tethered battery pack to handle the higher energy usage. Meta’s Quest line of devices are untethered to battery packs.

The AVP resembles the iPhone more than the MacBook (if we ignore the tethered battery pack). Eliminating controllers forces new user experience (UX) standards for app developers building for the AVP. This forces new UX standards for hand gestures only. This is the positive that comes with having no controllers.

There are also repercussions. A lack of controllers takes away precision user inputs.

Why are precision user inputs important? Games require very precise controls.

If history is any indication though, Apple doesn’t care much about gaming. Just compare gaming options on a Mac vs PC. Let’s go with the assumption that Apple is calling a mulligan on broader gaming market. This means the gaming market goes to Meta in the early years.

Why else might precision user inputs be important? The answer is design. Apple abandoned the gaming market but won the designer market with the Mac and MacBook. Walk through any Chelsea art gallery or architecture studio and you will only see Macs and MacBooks. This is the power of defining visual culture.

Designers and creatives are an under-analyzed market segment. Designer preferences lead to a trickle down effect in adoption rate.

MacBooks became ubiquitious because of designers indirectly. The Apple Vision Pro is not friendly to designers at the moment. Especially the price point in this economy.

The other thing standing in the way of adoption is battery life. The battery pack for the Apple Vision Pro only gives you two hours of battery life. Two hours of battery life is a fun demo’ing device. But it’s not a permanent replacement for current computers.

The iPhone had great battery life relative to comparable phone’s batteries. This is not true for the Apple Vision Pro yet.

The AVP is redefining graphics. The Apple brand allows Apple to use the best graphics hardware and sell at a higher price to consumers. Apple aspires to have the AVP replace all screens one day. It’s current commercial campaign shows the AVP being used for things laptops and TVs are typically used for. This is a strategy that realistically only Apple can pull off in the current market.

When you can sell a MacBook Pro stand for $1,000 - you can also add all the bells and whistles to sell a next generation HMD for $3,499.

Apple Vision Pro’s Lack of Apps and Developer Relations

The Apple Vision Pro reminds me of Nintendo’s N64 launch when it was going up against the Playstation and Saturn. The AVP is powerful device that lacks 3rd party developer support. In this analogy the Quest devices are the equivalent of the Playstation. This not a precise analogy as the graphics and hardware for the AVP are a higher but it gives an impression of what’s happening.

Two things stand out to me about Apple and third party developer relationships:

1. Non-existent Pre-launch Developer Relations - I don’t know of any independent XR studios that got early prototypes

2. App Store negative goodwill - The Epic Games legal situation (among others) has brought negative goodwill from App developers in general

Prior to the AVP launch, Apple had non existent developer support. QA testing that is only accessible in a few select cities is not developer friendly. Developer relations and developer experience were non priorities for the AVP launch. It is certainly weaker than all existing competitors.

Independent app developers are not incentivized to have Apple be a clear winner. Why not? It maintains the same paradigm as the mobile App Store market. Third party developers are incentivized to have a fractured market where Apple has less leverage than they have in the mobile market.

Apple at the moment is similar to the Microsoft of the late 90s in terms of developer brand perception. They are viewed as powerful but non considerate. When Microsoft made their entry into the mobile market, they had similar issues to Apple. No one wanted to go out of their way to help them win. I expect the lack of apps to continue for AVP until more devices are in the market.

Apple Vision Pro Launch Conclusions

Apple is positing itself as an augmented reality device. It is doing so because the AVP then becomes a computer with digital screens. The device is presented as a replacement for other computing devices and screens. It is using spatial computing branding to set itself apart from other devices in the category.

Other XR HMDs have emphasized experiences that are unique to XR HMDs. Apple is emphasizing what is familar. The purpose of a familiar but slightly better framework is that it changes the context from “do I like spatial computing / XR?” to “is this slightly better than other personal computing devices?”

The Apple Vision Pro is a a glimpse into the future XR HMDs shrink into the form factor of smart glasses and eliminates the need for other devices with screens.

It will likely do so with an initial content drought form 3rd party developers. The content drought is likely to continue in the near future though and the device will be anchored by internal 1st party apps from Apple. The lack of controllers also makes the device less friendly specifically to gaming developers that make games requiring precision controls. The AVP will also be moving forward without designers as an initial anchor userbase.

There are the challenges.

Apple also has the best brand in the world and lots of cash that it can throw around to solve roadblocks. This is a massive advantage.

Are there any topics you’d like me to explore about the AVP in the future? Let me know!

So I decided to ask ChatGPT a few questions.

Question 1: What are the odds that a new AI competitor led by Sam Altman and Greg Brockton would supersede OpenAI?

Tech figures on twitter believe it's an absolute certainty that a new Sam Altman would win. The odds form ChatGPT seem more reasonable to me. The overstating of the odds for Sam Altman is likely a product of being the previous president at YC. As a result, there is a large and loud group of supporters who are overestimating the importance of Sam Altman.

Question 2: What is the number of AI researchers with AI knowledge similar to Ilya Sutskever that could rebuild openAI from scratch given enough compute resources?

One point I think a lot of people are missing is the rarity of top tier AI researchers. There are not many people comparable to Ilya Sutskever. In some sense, these researchers are even more valuable than compute.

Question 3: How much money would a new Sam Altman and Greg Brockman startup need?

This estimate seems about right. It also speaks to the expense of creating a leading foundational AI company given the expenses associated with compute currently.

Question 4: How much did Cohere or Anthropic odds of superseding OpenAI increase after Sam Altman and Greg Brockman's exit from OpenAI?

People assume that an Altman and Brockman NewCo will be the default new victor and beneficiary. The biggest winners might be Cohere and Anthropic.

Question 5: What about Apple and Meta superseding OpenAI?

Meta and Apple are right around Cohere and Anthropic with Meta being especially competitive. As Meta and Apple get more data from spatial computing efforts and AI integrations. I personally think Meta’s odds are are a little better. Apple seems high but it also has nearly infinite financial resources to make up for in ChatGPT’s mind.

My final thoughts. If ChatGPT is accurate, we are looking at a relatively even six horse race at the moment at the foundational AI layer. Altman and Brockman’s odds are solid but no where near as certain as the tech elite make it out to be on twitter. Top experienced AI researchers are vastly undervalued at the moment given how few there are.

What are your thoughts?