my TL is full of BCI (particularly non invasive read/write), but no one with a concrete set of futures we can agree on i offer: emotion airdrop, the capacity to share feelings between friends, and for artists to craft multimedia experiences via "Palette", a service acting as a paintbrush for emotion. films should be able to make you feel things you would rarely reach yourself, like joy alongside melancholy, and accentuate certain emotions you are already feeling. to feel the deepest love, the greatest pain, and truly empathise with your peers and your culture. it is hard to imagine an in group/ out group dynamic surviving such close sharing of feeling. a sense of pushing against the edge of your mind, BCIs should not be predictive, they should require you actively "press into" the digital world, in a way that feels akin to a swipe on an iPhone. a simple action but one that is not made on your behalf, one that requires you make an active effort. if BCIs are predictive and decode your desires before you are aware of them, you won't have the sensation of having consciously requested something, moreso that it unearthed some desire that you may not even have yet been confident about: casting your thoughts into action too prematurely. the BCI cannot impinge on your sense of self and feel like a dark cloud inside of your finite consciousness, it should let you push into the digital abyss and expand your being. minimally invasive approaches that extend cognition, society has a natural disposition against plastic surgery and dispositions against invasive BCI, surgeries that would seem for vanity if purely for human-computer interaction. biohybrid BCIs give us the capacity to have brain systems that we "grow into" via axonal arborisation, making them societally easier to accept, and provide a read/write modality much denser than ultrasound based stimulation/readout. we need that density if we want to expand cognition, to provide a maximal bandwidth information channel into an external substrate that our consciousness fills. we're computationally limited. we need BCI to help unchain those biological limits. solving the coordination problem of society, rather than agents representing us, we can represent ourselves in every domain where we are required, outputting our thoughts at scales that would be unreachable via verbal communication alone. human thought can sit before a tool call, tool AI downstream of man without replacing human agency. that requires scaling cognition, scaling read-write throughout, and might be impossible without invasive BCI. but if executed well, could mean the resolution of conflict at a scale that is hard to imagine, the perfect transfer of information between one mind and another, including intuitional sensibility and that internal confidence one might feel around their views which only a lifetime of experience in a field can produce. seeing so much validity across heterogeneous views will enable true plurality in a way that we struggle with currently, where society wants a one-hot vector output of right/wrong over a floating-point value. if we can share enough, quickly and at scale, we might finally understand each other as a species. a lot of my recent thinking currently comes from discussions with @thryambakg on our drive a few weeks back from Santa Barbara to SF. there are more thoughts i will be sharing in the coming weeks on a canonical vision for the future and maintaining human agency within it. we must rage against those who wake up without a love for their lives, and force them to live their lives fully. to choose for themselves what they want in existence. to prevent algorithms from deciding the fate of humanity. we must build a future organised around pointing to the Heavens, stressing a correct direction of pursuit over avoiding fates. we will never win by being EAs scared of downside, we must select the outcomes we wish for.

today we're launching Frontier Computing (YC S26) we grow neurons that colocate memory and compute, so you can more easily train ML models. here's one of ours playing Frogger. and we're building a 500M-neuron cluster. live EOY 2026. that's 2,500x larger than the major biocompute systems available today and we'll keep scaling until large model training on biocompute is widely accessible. after all, human intelligence is built on a lipid, amino acid, glucose and vitamin cofactor soup: eating food is less costly than eating GPU compute