

✓A seed-stage venture firm investing in open networks. twitter photo / background: @artbydouglxss


Proud to work with @vincentweisser, @willcb, @caoshuxun and the rest of the @PrimeIntellect team. They build open frontier models and the infra to post-train them, and went from 11B to 1T+ monthly tokens on OpenRouter in under a year!

Fountain x DATALAND Fountain Coffee Club’s final stop of the year takes us to @refikanadol's DATALAND on December 10. Link to RSVP is below. Internationally recognized for transforming vast datasets into immersive artworks, Anadol continues to push the boundaries of art and technology at this new museum. DATALAND is a pioneering new addition to Los Angeles’s cultural landscape. Dedicated to AI art, the museum features five multi-sensory galleries where artworks can respond to visitors and evolve in real time. We couldn’t imagine a better way to close out the year; we hope you can join us.


"We are at the beginning of a long arc for how AI will interact with the web at scale, and the subsequent economic disruption that comes with it. It will break content publishers’ business models as the web shifts towards agents. It will erode the margins of businesses that profit from friction, shifting power back to consumers. It will create opportunities for agents and infrastructure providers, alongside new business models in content and commerce."

Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values. Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line. After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs. The values in an open model are one fine-tune away. Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete

Brokered by Fountain ⛲️: “Saturn” and “Jupiter” by @summergwagner Summer describes how each piece anchors the Parody of a Tangled Thread series: “Saturn opens the collection, the calculated restraint of the celestial discipline as the figure looks over the vast, star-lit water. Jupiter, at the end of the collection, is that same figure’s embrace of expansion, release. The two together express the binary of direct experience and reflection.” Congratulations to the new collector!

We built Atlas to make professional 3D production faster and more reliable for studios, while driving adoption for 404. One example is Parallel Studios, which is using Atlas inside Colony, its AI-powered sci-fi game built in partnership with Google Cloud. Players can describe the gear they want, and Atlas turns those ideas into on-style, game-ready 3D assets in real time. From prompt to image to 3D, it’s a glimpse at a new era of AI-powered games.

my guy @alexpodkin is cutting metal and making gears right here in los angeles county as of twenty minutes ago let's bring our manufacturing networks back home @gearsdotco

As AI masters turning information into useful outputs, the bottleneck is AI accessing that information in the first place. This is the most fundamental yet least understood part of the "agentic" stack, and Ronan does a great job breaking down this budding economy in his piece.

My new piece for @collab_currency Agents are rapidly changing how the web is consumed and who profits from it, rewriting the economic bargains that have sustained much of the open web for thirty years.

as the world moves from single-player computer use into our new agentic paradigm our legacy business models for the internet will too evolve a timely essay exploring these new dynamics from @ronangmi for @collab_currency

roughly two years ago @artonblockchain shared an idea with me for an onchain generative artwork it was titled: LIFT (a self portrait) the collection was deployed on @ethereum in august 2025 powered by @artblocks_io contracts LIFT (a self portrait) consisted of 1,000 procedurally generated, fully onchain artworks and video games, each playable through a Game Boy Advance emulator or physical cartridge at the center of each game was a character called 'pixel man' a procedurally generated self-portrait of snowfro who was asked to lift a heavy boulder with the assistance of his onchain friends, while overcoming obstacles and opposing forces over 300+ artists and friends of @artonblockchain were invited to create onchain pixel supporters for the project, which populate across all the 1,000 unique artworks importantly the network snowfro built around LIFT (a self portrait) was always meant to be a starting point the beginning snowfro had a much larger network he wanted to build designed around the second chapter of his LIFT universe tonight in marfa TX our team at @glitchmarfa is honored to present snowfro's second chapter in partnership with @LACMA and sponsored by @Fountainxyz the new collection is called LIFT (together) LIFT (together) is an interactive artwork and system designed by Snowfro combining a physical pedestal, offchain software, the @ethereum blockchain, and a global competition the system invites anyone to walk up to the pedestal design their very own pixel supporter engage in a game of speed and help lift snowfro’s rock upon completion the system generates a 1/1/x onchain exhibition poster on @ethereum each onchain poster is also unique artwork, incorporating your designed pixel character, procedurally generated assets, and elements using @ethereum entropy a small number of artworks in the collection may also find *rare* and *very rare* artwork attributes each exhibition poster is also a record of location, printed on each artwork over the next decade the LIFT (together) pedestal will travel across the world allowing collectors across different geographies and institutions to join snowfro’s onchain network and tonight, the journey begins in marfa, TX join us at @glitchmarfa ❤️ gm

actual computer turns your local computers into a department- scale AI factory @actualinc

a few notes on local AI and our computers that run it around fifteen months ago, i led the pre-seed round for a weird local AI project it was called @actualinc led by @Tom_A_Lynch for the first time in the AI movement open models are now good enough for a large and growing set of workflows, especially where latency, privacy, bandwidth, and token cost are critical hermes agent from @nousresearch and ‘lab’ by @primeintellect have broken this idea wide open with killer products today, built on open rails we are no longer forced to assume that every useful task must route through one of a few centralized frontier APIs and so tom and the squad at @actualinc have painstakingly created a new type of proprietary inference engine for all types of computers maybe you are an owner of one of the tens of millions Apple macs or an owner of one of the millions of RTX-class machines or maybe you run a small team or company and you are starting to realize what many others across the world, have started to realize which is that sometimes centralized AI systems bring their own set of unique problems especially as our workflows move into an always-on, agentic era which might sometimes come with: - exorbitant costs - no privacy - lack of control but a problem remains running local AI is still too hard firing up one model on one machine is difficult enough but running models across different machines means navigating drivers, networking, model sizing, memory limits, API setups stitching together hardware is messy across Mac, Windows, Linux, NVIDIA, Apple, AMD heterogenous systems were never meant to work seamlessly together like this today, you'll find that most of the local inference software available is built around single users, for a single machine use case running inference capacity in bulk, across many users, and many kinds of computers that is a different animal altogether tools like llama.cpp made local models accessible, but they are not designed to coordinate tens of thousands of heterogeneous machines under a live supply network vLLM-style batching can be economical, but vLLM is limited in the machines it can run on the local AI movement requires a simple product experience, for a tough engineering problem that is the problem @Tom_A_Lynch pitched me fifteen months ago and the problem @actualinc has been engineered to solve now with the help of his cracked and growing team @skooookum, @somewheresy, @jack15032, and the rest of the crew starting with their first product go to actual.computer in under one minute, install the @actualinc client, authorize your machine, load a model, and @actualinc turns your personal device into a private inference endpoint in less than twenty seconds, add a second, or third, or fourth machine of any type and it will start behaving like a small private cluster from a technical POV actual runs as a lightweight daemon on each machine and turns local hardware into a routable inference endpoint each node reports its available memory, accelerator type, model state, and health back into the cluster, while the user interacts with a single API surface instead of managing ports, drivers, or SSH sessions a relay layer lets requests reach machines without forcing users to configure networking or expose their devices directly and as a result computer owners of all kinds of computers can get local inference clusters - when they want to control costs - when they want to control privacy - when they want to build on top of a neutral stack eventually, over time computer owners running these local clusters should be able to seamlessly contribute their unused capacity back to the @actualinc network for a market to price and for someone else to purchase you can think of @actualinc as building the engine and control plane for datacenter style throughput in homes, in offices, in small to medium sized businesses, and in very big sized businesses because there are hundreds of gigawatts already on the grid right now, that are waiting on desks, in storage closets to become actual computers

since 2020, one of my north stars is that the best and most robust networks will be built around: - openness - credible neutrality - user sovereignty by many metrics, @nousresearch hermes agent is now the most widely used open source AI agent in the world and notably the value of hermes agent is not just that it is growing incredibly fast it is that it is growing incredibly fast in the right ways devs are cloning it contributors are improving it model providers can route any/all intelligence through it and more recently enterprises now feel safe to build on top of it in a world where synthetic intelligence is one of the most important inputs into labor, software, hardware, creativity, decision making it is more important than ever that our infrastructure of the future should not be a black box gatekept by a few to create distribution, pricing, or performance advantages at the expense of the rest of us it should be open, auditable, extensible and it should be controlled and operated by the people and organizations that use it proud friend and long time investor of this mission driven crew at @collab_currency keep up the people's work @dillonrolnick @karan4d @Teknium @theemozilla @SHL0MS @yeahfortommy @gradyacc and the rest of the killer squad at @NousResearch ♥️