Training strong AI models outside large data centers requires solving a hard coordination problem between distributed GPUs.
@chutes_ai just announced they trained a recurrent model with Parallax across distributed GPUs in a fully non-blocking training setup, staying within only a 0.6% quality gap versus centralized training.
A fully non-blocking setup means each GPU keeps training instead of pausing until every other GPU has finished synchronizing.
That matters because recurrent models are harder to split across many GPUs than transformers, which makes this a strong test case for decentralized training.
Another concrete milestone for decentralized AI training from a Bittensor subnet.
Vincent - we are building our own company “Brain” for our vertically integrated real estate services business across management, brokerage, maintenance and construction divisions. We are currently running a Hermes agent on open source models through open router but i am interested in the cost of running our models locally so we own our own intelligence full stack.
Who can I talk to at Prime Intellect about this? Very interested in what your team has built, congrats on the series A!
We raised $130M @ $1B for our series A
To build the open superintelligence stack for everyone
Pre-training concentrated frontier AI in a handful of labs. RL changes who can build frontier AI and just works across almost any verifiable domain. We want to enable everyone to train their own agents.
Companies can now own their model optimization loop: train directly on your product, optimize for your specific workflows, and build agents that improve continuously in production
Owning this model <> product improvement loop is how you build a compounding moat in the agentic era
Super grateful to serve over 6k+ customers, including many leading AI startups, neolabs and enterprises already building on our stack, and to our incredible team for shipping hardcore!
We train open frontier models and ship the same stack to our customers. Its spans the full stack of training, deploying and continuously improving models — compute, large-scale RL, environments, sandboxes, evals, and deployment.
We're excited to be joined by angels who are building the frontier themselves, many of whom we work closely with:
@johnschulman2 (Thinking Machines), @dwarkesh_sp, @AravSrinivas (Perplexity), @karimatiyeh (Ramp), @levie (Box), @_milankovac_ (Tesla), @winstonweinberg (Harvey), @amspector100 (Flapping Airplanes), @jeffwang (Cognition), @_arohan_ (Core Automation), @marksaroufim (Core Automation), @mikeknoop (Zapier, Ndea), @eastdakota (Cloudflare), @BrendanFoody (Mercor), @devanshpandey (Standard Intelligence), @hwchase17 (Langchain), @nicoup (Fleet) and many more
We're a small team building open superintelligence
> Reach out if you want to partner training, deploying and continuously improving your own frontier models for your use case
> Join us to build open superintelligence — we're hiring across all roles including RL, inference, distributed systems, full stack engineering and compute.
$NVDA being in on this is big time for open source intelligence.
Open source, distributed AI is the future or else we are not sovereign as individuals.
I wrote an essay on why open source AI is the way forward here: https://t.co/vaZZYBANQC
Announcing our $130M Series A to build the Open Superintelligence Stack
Led by Radical Ventures, with NVIDIA, Intel Capital, Dell Capital, and existing investors
Train, deploy, and continuously improve your own models using our stack.
Own your intelligence.
The market is approaching a generational long opportunity on $TAO.
My target for the upcoming cycle is $3,000 w/ an ultra bull case between $5,000 and $7,500
Sub $200 is a gift. Few.
Jefferies: AI Crescendo
> Massive Spending Forecasts: The four major US hyperscalers (Alphabet, Amazon, Meta, and Microsoft) are expected to spend roughly US$700bn in capex this year and over US$800bn next year.
> Macro Impact: If you include Oracle, Anthropic, OpenAI, and neo-clouds, total capex is estimated to surpass US$1tn next year—equivalent to roughly 3% of US GDP and 33% of total pre-tax profits for all US non-financial companies.
> Cash Flow Strain: This aggressive spending is rapidly eating into cash flows; capex as a percentage of operating cash flow for the four major hyperscalers is projected to skyrocket to 92% in 2026, up from 41% in 2023.
> Bond Issuance: The four major US hyperscalers have issued US$144bn worth of bonds so far this year, compared with US$83bn in 2025 , according to Bloomberg. Alphabet, Amazon and Meta issued US$52bn, US$67bn and US$25bn in 1H26, while Microsoft has not yet issued any bonds. In addition, Oracle has also issued US$25bn of bonds year-to-date.
> Memory Dominance: There are now only three dominant suppliers of DRAMs globally, compared with 12 prior to 2012, with an estimated market share of 89%.
> AI Driving US GDP: The US economy is highly dependent on this trend; real AI-related capex alone contributed 1.13 percentage points (or 42%) to the US real GDP growth of 2.68% YoY in 1Q26.
US mortgage rates are choking the housing market:
The 30-year fixed sits at 6.43%, while the average rate on existing mortgages is just 4.4%.
Homeowners won't trade a 4% loan for a 6% one. Buyers can't afford to step in.
Existing home sales are running near a 4.17 million annual pace, close to the slowest in three decades.
More than half of active listings have sat on the market for over 60 days.
The median existing home sold for $429,300 in May, an all-time high.
New single-family sales fell -7.3% in May, pushing new home supply up a full month to 10.3 months, the highest since 2009.
Elevated rates and record prices are keeping buyers on the sidelines.
We have a commercial construction and brokerage division and both are much slower at this time vs. last year. Capex decisions seem to be getting put on hold right now.
Our property management division is also leasing at a lower velocity. Most of the softness in 80-100% AMI bucket. Above 100% AMI still holding up pretty solid. Based in NC.
I spoke to McConnell for about 20 minutes this morning.
He said we should end the war with Iran, quit giving aid to Israel, stop spying on Americans without a warrant, and he’s really sorry about how my primary turned out.