posts about ai, games, startups. vp @theresidency, prev co-founder, ErthAI; GM, @Zynga; team Workplay Ventures, ReinventCap
San Francisco, CA
Joined August 2008
- Tweets4.3K
- Following791
- Followers3.7K
- Likes20.3K
Dan Garon retweeted
this account is decompiling a Samsung smart tv and it's actually worse than I was expecting, and I was expecting the worst
this is the most bullish post I’ve seen for AI founders. study these charts. it doesn’t foretell a bubble, it suggests that the AI boom is barely in its first innings.
we are *so* early. AI has permeated almost nothing outside of software engineering. There are no measurable productivity gains in the federal numbers — incomprehensible to anyone in San Francisco.
you think token consumption is parabolic today? guess what, nobody is burning tokens yet outside of the top 5% of companies, and of them, vast majority still think AI is better google search. nobody outside of your x bubble understands how an agent works or why Sam Altman says our systems are not patchable.
soaring interest rates will not burst this bubble — at least not for while. Rising rates will crush margins on everyone outside of tech, and those companies will realize they too need to start burning tokens. They will turn to tech for help.
The anti-AI games post from angry gamers are frustratingly naive. Every one of your favorite games is derivative of a previous game! The greatest game designers crib from the top charts.
Shariki was probably the first match 3 game, created by Russian programmer Eugene Alemzhin in 1988. Popcap added jewel fiction and it became Bejeweled, which EA bought, and that was forked into Candy Crush Saga and Puzzle & Dragons. Each new generation generated massively more revenue than the prior. Interestingly, there are roots in matching rock games going back to Roman emperor Hadrian.
Tetris came out in 1984 (there’s an awesome Apple TV movie on this, wild story) and sold almost 200 million copies. It was endlessly copied and sparked a handful of puzzle genres: Swapping became merging, balls and blocks became jewels, candy, and bubbles. Progression maps and character combat showed up in some variants. Like in match 3, each generation became more fun and more profitable.
The history of game making is reverse engineering and so is its future. And that future is bright.
I pointed Claude at the below post and it embedded all 146,060 Steam games, sampled 320,000 points between them, ranked them by how far they land from anything that exists, and used creative judgment to pick a winner... and it cooked!
It's pick: Short Position, a co-op factory tycoon where one player has secretly shorted the stock.
4–8 players run one industrial empire: mines, smelters, rail, contracts. There’s no kill button and no sabotage minigame. The traitor plays the factory game badly on purpose. Overbuild a warehouse, route ore to the wrong smelter, sign a contract with a penalty clause on page 4. Every move looks like an honest mistake.
The twist is that suspicion is expensive. Anyone can audit the production logs, but audits cost money and spook investors. The traitor doesn’t need to wreck the factory. They just need everyone paranoid enough to do it themselves.
Each quarter ends with an earnings call and a board vote. Fire an honest exec and you lose their whole department. Four quarters, ~40 minutes.
It sits farther from its nearest neighbors than 83% of real Steam games. Factory games have co-op, social deduction games have tasks, and nobody has made the work itself the place where you lie.
Walking speed = Raw intellectual capacity.
Do your candidate interviews outdoors and measure their walk speed?
As an aside, IQ strongly correlates with walking speed.
The correlation is, like, extremely significant. There are essentially zero outliers.
Top quintile walkers are a full standard deviation smarter than bottom quintile walkers.
Readers added context they thought people might want to know
A 2019 Dunedin study reports moderate correlation (β=0.38) between midlife IQ and gait speed with ~16-point mean quintile difference. "Essentially zero outliers" does not match this association's strength.
jamanetwork.com/journals/jaman…
The executive who ran Minecraft for 9 years — Helen Chaing — was passed over for the XBOX CEO role and is leaving. She ran Microsoft’s only moddable franchise and has been at the company for more than 20 years. I’ve only met her a few times, so don’t know her personally, but she has the most insane resume.
Aisha is rebuilding the entire Xbox strategy and the old guard is rotating out. (She had been COO for 3 months, likely as part of Aisha’s transition).
Helen is likely the most sought after executive in all of gaming right now. Will be curious where she lands next. Congrats on a generational run.
They’re all similar, all feel mostly like OpenClaw forks. None are running the most powerful models, but it doesn’t seem to matter. Nobody pushing any limits yet. Claude is probably best positioned for heavy work, Muse for life admin. I don’t know anyone using Instinct anymore.
Muse = fastest and cheapest; best for computer use, ability to scrape Meta ecosystem. They are teeing up this being the ultimate commerce agent, and that’s plausible.
Grok Bot = feels similar to Muse, but more capable in enterprises, and has full access to X. Creating multiple bots with context is better than Muse side chats.
Instinct = seems to be trained around travel use cases, otherwise feels the least capable. Until the move away from SMS interface, it’s not a serious contender for heavy work. Big trust issues given the context they need and the “move fast and break things” approach.
Dots = Feels like you’re using codex with a virtual machine. They have not invested in any UI/UX innovation and it shows. As it stands today, most users won’t notice any difference than just using ChatGPT. OpenAI should be years ahead of everyone given how stacked their team is and their strength in consumer, but it’s not showing up yet. Codex running locally is still a better computer use model than anything in the cloud because it won’t get blocked by default.
Duet = Still very early, but their team sees to be building Slack, except every channel is an agent. They’ve also introduced a “Make” tab which integrates Claude Design. As it stands today, it’s not useful, but it’s pointing to some very powerful primitives. I suspect they are about to set a new bar for agents.
The agent wars have begun, warning shots have been fired, but nobody has shipped any of their biggest ideas.
This is the future of how small businesses will be able to talk to their customers and interact with their own tools.
Block is building the entire stack for commerce on Main Street, from AI kiosks and agentic payment rails to voice models that run on edge devices. Very cool.
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
This is so unreasonably cool. Whoop for whole body movement.
Next stop - I wore a motion capture suit across a beginner Hyrox simulation, and got very cool data out of this.
I did this at @7x7 and no one got to know except the founder. I did this at one of their Open Gym sessions.
What do you all think of the interface and my {movement}
First reaction on Claude Duet is that it’s setting a new bar for agent UI/UX. Have been playing with the TestFlight all morning.
— Dedicated app that isn’t Claude (OpenAI Dots still buried in ChatGPT. Muse, Grok Bot are dedicated apps.
— Fully integrated Claude design / maker space. This is the best I’ve seen in any agent tool by far
— Slack-like channels that allow for meta prompts. Grok Bot is similar, Muse still messy on this front. Instinct is obviously only single threaded and that’s a huge disadvantage for them. I can imagine these becoming “multiplayer” relatively quickly.
— Task list that applies per channel. This is wonderful!!
It’s not obvious that Duet has much more capability than its peers, but the interface seems to enable better productivity. I expect everyone else to converge on these patterns quickly. It’s just too good.
The @eriktorenberg Oscars!
Can’t wait
Excited to announce the Cosign Awards: celebrating exceptional people, companies, and products that had a massive impact in 2026.
cosign.co/awards
Anyone can nominate and vote. Award winners will be chosen by the public in conjunction with a panel of judges, including industry leaders like @lennysan, @packyM, @amasad, and many more. We’ll host an event in December honoring the winners (10 per category).
For people categories, we’re mostly focused on everyone except the CEO–everyone else building the company behind the scenes.
Companies with the most amount of people nominations will be listed on our “Top Companies to Work For” page.
True to form, at Product Hunt I helped launch the Golden Kitty awards a decade ago to recognize great products. Soon after I launched the Rise Awards, honoring executives and ICs building companies that make them. The Cosign Awards is the next iteration.
We have 41 categories across people, companies, and products.
People categories:
- Future Founders
- Rising Stars
- Engineering
- FDE
- Product
- Design
- AI Researchers
- Marketing
- Comms
- Growth
- Chief of Staff
- Operations
- Talent
- Finance
- Public Intellectuals
- Podcaster of the Year
- Newsletter Author of the Year
- Poster of the Year
Company categories:
- Rising Stars
- Best Place to Work
- Founder Mafia
- Talent Density
- Highest Aura
Product categories:
- Product of the Year
- Launch of the Year
- AI Assistant
- AI Language Model
- AI Coding
- Dev Tools
- Open Source
- Consumer
- Productivity
- Creative & Media Tools
- Product Design & UX
- Enterprise B2B
- Hardware
- Robotics
- Self-Driving
- Healthcare
- Vertical Fintech
- Legal
I think large part of the cambrian explosion we are about to see in precision medicine is simply a result of shortening the feedback loops between intervention and diagnostics.
for example, if you can measure metabolites in a tumor on a daily basis, you can much more quickly iterate between treatments.
this is all exciting news, and congratulations to this team.
We’ve raised a $45m Series A, co-led by
@foundersfund and The @OpenAI Startup Fund, to build cancer therapy that learns. Technology is rapidly changing medicine, but data loops and rapid adjustments are still largely missing. When you start cancer treatment, knowing whether what you’re doing is working is too slow. Clearly there is a better way. Build systems to watch cancer in real time, then feed that data into a system that itself can treat the disease. This seeds a future where, unlike drugs, we develop medicine that continuously learns to improve itself. We've been building neurotechnology and BCI platforms to do just that.
@coherenceneuro
- now may literally be the worst time in 40 years to buy a home. sorry millennials.
- housing market is seeing the most severe collapse in transactions in modern history, but without a drop in prices.
- this the lowest reading since the housing index began in 2001. Includes 2008 financial crisis / COVID
- closest comp is the interest-rate shock of 1979-1982. Home prices similarly didn’t drop in nominal dollars, but home values were eaten up by inflation. Buyers got screwed then too.
Dan Garon retweeted
Simple: ship a working version of the @theranos vision powered by SI
Could be 10% of the original vision, with the promise to keep iterating
For bonus points open source all the technology and patents
And given the pace of super intelligence, someone will and then folks will start saying “seeeeeeeee! @ElizabethHolmes was just too early!!!”
I will leave it to @nathanfielder and @JohnCarreyrou to adjudicate that moment! 😂😂😂
[ go see the film ]
Congratulations! you’ve been hired by Billy Evans and Liz Holmes to handle their PR.
What do you advise them? Options below (collected from my various sf tech group chats). What did I miss?
1) Deflect.
- Issue a short, timid statement
- reiterate her innocence, acknowledge mistakes
- suggest the filmmakers misled the audience / the family
- ask for privacy for the kids
- then lay low.
Most popular recommendation.
2) Disappear.
- Decline press,
- stop posting from Holmes’ X
- avoid any public fights with Fielder.
- Anything they say will be used against them
- The time to reappear in public, if ever, is after Holmes is out of prison and the public has moved on.
Second most common recommendation.
3) Mea Culpa.
- solely Holmes, no Evans
- accept substantial responsibility
- a chance to earn sympathy / reframe the damaging parts of the story. (“I wasn’t fully honest with our wealthy investors, but I feel so bad I hurt the people who trusted me” kinda thing).
- vibe is she has learned and matured; deep voice is gone.
- Focus on humanizing, including that she was in her early 20s and clueless.
- narrative is second chances + overly zealous prosecution
- Hard to pull this off logistically if she can’t do an interview from prison; may need to wait until she is in a halfway house next year. Nobody thought either of them would actually go for this.
4) Full Trump.
- Come out SWINGING.
- Focus is on Evans, who arguably has a much bigger reputational problem than Holmes
- Produce LOTS of evidence, including trial defense. show this case was less clear than everyone thinks.
- publicly confront the filmmakers
- truly “own” Holmes’ and Evans’ roles as the villains
- press on that Fiedler has a history of tricking people, and say they were tricked.
- framing as ambitious entrepreneurs with powerful enemies, eg powerful lab diagnostics oligopoly they are taking on.
- Compare Holmes to other founders who were doubted but overcame. (Evans somewhat adopted this in the film when he spoke of his vengeance and powerful enemies.)
For this to work, Evan needs to be so relentless that he earns support for his grit and chutzpah.
Congratulations! you’ve been hired by Billy Evans and Liz Holmes to handle their PR.
What do you advise them? Options below (collected from my various sf tech group chats). What did I miss?
1) Deflect.
- Issue a short, timid statement
- reiterate her innocence, acknowledge mistakes
- suggest the filmmakers misled the audience / the family
- ask for privacy for the kids
- then lay low.
Most popular recommendation.
2) Disappear.
- Decline press,
- stop posting from Holmes’ X
- avoid any public fights with Fielder.
- Anything they say will be used against them
- The time to reappear in public, if ever, is after Holmes is out of prison and the public has moved on.
Second most common recommendation.
3) Mea Culpa.
- solely Holmes, no Evans
- accept substantial responsibility
- a chance to earn sympathy / reframe the damaging parts of the story. (“I wasn’t fully honest with our wealthy investors, but I feel so bad I hurt the people who trusted me” kinda thing).
- vibe is she has learned and matured; deep voice is gone.
- Focus on humanizing, including that she was in her early 20s and clueless.
- narrative is second chances + overly zealous prosecution
- Hard to pull this off logistically if she can’t do an interview from prison; may need to wait until she is in a halfway house next year. Nobody thought either of them would actually go for this.
4) Full Trump.
- Come out SWINGING.
- Focus is on Evans, who arguably has a much bigger reputational problem than Holmes
- Produce LOTS of evidence, including trial defense. show this case was less clear than everyone thinks.
- publicly confront the filmmakers
- truly “own” Holmes’ and Evans’ roles as the villains
- press on that Fiedler has a history of tricking people, and say they were tricked.
- framing as ambitious entrepreneurs with powerful enemies, eg powerful lab diagnostics oligopoly they are taking on.
- Compare Holmes to other founders who were doubted but overcame. (Evans somewhat adopted this in the film when he spoke of his vengeance and powerful enemies.)
For this to work, Evan needs to be so relentless that he earns support for his grit and chutzpah.
This is so stunning, undeniably be the most insane example of what is coming in the next decade of bio.
- Analyst knows nothing about biotech, but intrigued by the hype
- Gets access to unreleased Claude model, uses it to run an autonomous drug design campaign to create antivenom for scorpions
- $200k budget; 25% compute, 75% for the wet lab
- 3 day design time; 35 page prompt, generated mostly by the model itself. only FIVE human messages were sent to Claude
- Claude could answer its questions by emailing vendors, which it did!
- 300+ agent swam designs, scores, and delivers a set of proteins, then sends to a wet lab for testing
- 211 protein designs tested, of which 17 bind to the target in four of five toxins
- Compresses years of biology work into a few days
- All done by a team that knows virtually nothing about life sciences
What is AI really capable of in bio?
To separate hype from reality, we spent 3 months working firsthand with @AnthropicAI's internal research model to see if it could create an experimental scorpion antivenom.
Introducing Citrini Therapeutics.
citriniresearch.com/p/citrin…
