Ideas worth your time
Discover ideas, people, and practical knowledge worth exploring.
Explore
Report · Money
There's something like 4 billion people in the world who don't have access to any brokerage or US investment account — the access gap tokenization and stablecoins address.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
moderate
Recommendation · Money
The speakers suggest exploring photonics as a 'why now' technological substrate for novel compute models, resistive memories and memristors as exotic devices with interesting new physics, and coupled oscillators as circuit ideas with nonlinear analog properties.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Recommendation from Y Combinator Paper Club
exploratory suggestions, not demonstrated results
Recommendation · Money
The discussion proposes taking inspiration from the brain's continuous sensing to build edge devices that sense, react, and adapt in real time, and explores event-driven computation the way the brain works. It also raises on-device learning so a network can self-improve based on what it perceives, and spiking networks as a way to train neural nets on spikes rather than conventionally.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Recommendation from Y Combinator Paper Club
framed as exploratory questions, not established results
Recommendation · Money
Armstrong recommends that companies that started AI adoption with table stakes — coding tools, data-repository integration, ML for fraud/risk prevention, and AI-handled customer support — should then move toward recursive self-improvement systems.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Recommendation from Brian Armstrong
Report · Money
The initial Doom approach used a closed-loop PPO architecture: the encoder passes the image through a CNN, concatenates scalar observations, passes to an MLP, and outputs frequency and amplitude per channel (in the open-source code it did not pick the channel); the decoder uses a linear readout; the PPO critic keeps feedback to the cells stable; the loop updates every ~2,000 steps.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Alok
explicit
Creator perspective · Money
The guest explains that unlike reservoir computing, the culture changes: sending a given frequency and amplitude changes the culture differently every time, so you don't know how it will perform. Pong is simple enough that encoding and decoding can be hard-coded by hand.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from unknown
Creator perspective · Money
The reason the team took an application-specific, fixed-weight approach is that it is very expensive to rewrite or keep the weights in the optical domain.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Ilker
endorsed
Creator perspective · Money
The next milestone for making optical computing practical is demonstrating an end-to-end system at very large scale, like a billion-parameter scale model, so energy and latency advantages survive when systems are scaled out.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Ilker
endorsed
Creator perspective · Money
A traditional computer architect asks about thoughts on optics-based storage, noting that if optics-based storage existed, one could bypass the ADC entirely. Optics-based storage is one of the biggest bottlenecks of the technology, per the presenter. Optics (as in CDs, DVDs) is a great storage medium for archival or long-term storage and works quite well, with many different ways of doing it. The challenge with optical storage is how to quickly and cheaply write and read memory in the optical domain. Historically there have been many attempts at optical memory, including holographic memories and photorefractive materials, but so far none has been as cheap and accessible as electronics. The reason electronics stayed as the main medium of compute is that optical memory attempts have not been
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Ilker
described
Report · Money
As of 2026, neuromorphic computing remains squarely in R&D land, with the field working on matching biological inspiration to physical limitations.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Alok
Creator perspective · Money
When scaling the number of diffractive layers or changing the number of pixels per layer, the optical system shows power-law scaling behavior similar to digital neural networks. The team expects these optical models to scale in a similar way to digital models as more parameters are added. Adding extra parameters to the optical system improves image quality performance in an expectable way. The team showed first premises that when more parameters are added, the optical models are expected to perform better, as in the digital neural network case.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Ilker
described
Report · Money
In the Doom setup, communication with brain cells is primarily via electrical stimuli: game state variables are encoded into stimuli (channel, amplitude, frequency) on a 2D multi-electrode array; the neurons accumulate charge and spike like leaky integrate-and-fire neurons; and spikes are decoded into a specific game action.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Alok
Principle · Money
Armstrong argues everybody should have access to an AI financial advisor now, the way wealthy people have investment advisors — and Coinbase has launched an advisor in the Coinbase product.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
Principle · Money
Armstrong frames cognitive enhancement as a race: AI is getting smarter and smarter, and humans need to at some point take control of our evolution and get smarter and smarter. The host frames it as how to increase the collective intelligence of humanity. Armstrong notes that in development of children, giving them enough folic acid and certain nutrients in certain countries actually raises IQ — they just develop better — and says he might host a dinner on cognitive enhancement, exploring ways to target therapies in adults, children, or embryos to level the playing field of baseline cognition and IQ that people are born with.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
moderate
Report · Money
Armstrong invested in a company called Preventive that's doing embryo editing, which is a controversial topic — people have seen movies like Gattaca. According to Pew Research, about 80% of Americans support embryo editing for disease prevention — nobody wants rare genetic diseases for their children; everybody wants healthier, better lives for their children.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
Creator perspective · Money
Armstrong states his main focus going forward is agentic finance at Coinbase, in a world where AI is getting smarter and there will probably be more AI agents than humans in the not-too-distant future. He argues humans still need to advance and progress even then.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
The 'more agents than humans' claim is hedged ('probably').
Creator perspective · Money
AI agents today get stuck at paywalls and resource sign-ups (credit cards, AWS) because they lack their own spend accounts; providing agents with funds removes this bottleneck.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
Report · Money
AI agents doing work for users often get stuck at paywalls and credit card prompts, which motivates agentic-native payment rails.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
Principle · Money
Armstrong finds it fun and intellectually stimulating to create companies that can advance the world, but warns that trying to do multiple things too early is dangerous because the distraction can be really dangerous and arguably irresponsible. He cites Joe Lonsdale, Sam Altman, Palmer Luckey, Elon Musk, and Daniel Lech as examples of people who have successfully incubated multiple companies. He notes that at a certain size of company, a CEO and a pure investor are both capital allocators — basically just allocating budgets and capital — and that for a founder incubating companies, time is probably the most valuable part of what they contribute, more than capital.
No Priors: Artificial Intelligence | Technology | Startups · Coinbase’s Everything Exchange: Agentic Finance, Stablecoins, and Tokenization with CEO Brian Armstrong
Claim from Brian Armstrong
Prediction · Money
The host predicts that in 10 years, we will stop using backpropagation.
Y Combinator · What If We Stopped Using GPUs? | YC Paper Club
Claim from Y Combinator host
explicit prediction, not a fact
Explore
Published routines with a defined commitment. Open one to inspect it; adding it to your plan happens there, never here.
No published protocol matches this selection.
Explore
One page per source: the complete extraction, its evidence, and every cleared result in context.
Source analysis · Money
Y Combinator · 177 cleared results
Source analysis · Money
No Priors: Artificial Intelligence | Technology | Startups · 154 cleared results
Source analysis
PowerfulJRE · 0 cleared results
The creators behind most of what cleared.
177 cleared insights match
154 cleared insights match
Choose sources → to get a library of your own and a daily brief.