Dad had a house worth $1.1M He sold it He put $1.1M into Treasuries at 5.1% $56K a year covers his rent $1.1M is still there Renting is no longer throwing money away Owning a house is now a losing trade

Really disappointed to see this. Here are the facts: -You did not terminate me. I resigned from my advisor position on Monday and told you I was going to Cognition. In response, you asked me to consider a full time role at Factory. I told you I wasn’t interested. -You claim that I sat through a board meeting while ‘confiding’ in executives at Cognition. The last board meeting I attended was weeks before I had ever spoken to Cognition. -Cognition has never asked about information on Factory, nor would I ever give it.

aws made the industry soft. we're used to them handling all the hard problems and everyone can be a pure product company with 85% margins ai infra is not mature enough for this but everyone is pretending like they can offload the actual work to someone else it's why there's 27 model router products. even most inference companies do the bare minimum on compute someone else set up and it all feels successful because there's so much money around it's hard to justify doing the hard thing when the easy thing is profitable but the gap between what openai, anthropic, deepseek, moonshot and everyone else is huge and it's easy to forget that someone actually has to do the work to close it

We have clearly benefited a ton from the ecosystem of ever-improving open models and the published research behind them. We will contribute back in a major way.

Three new DSpark speculators on the Red Hat AI Hugging Face, for faster inference through parallel drafting: - Qwen3.6-35B-A3B - gemma-4-31B - GLM-5.2 (updated) DSpark drafts a whole block of tokens in one pass, then verifies losslessly. On GLM-5.2 that's 4.0x faster decode, 137 to 551 tokens/sec single-request, and 1.57x higher peak throughput (1,402 to 2,203 tokens/sec) with an acceptance length of 5.47 on math reasoning. Serve them with @vllm_project. Links below.

More types added to the Thinking Orbs library. Large and small sizes, light & dark mode, playground, zero dep orbs.jakubantalik.com npm install thinking-orbs Huge shoutout to @a_brinza who also worked on this

We're excited to welcome @makora_ai to Infron 🎉! Makora automatically writes and tunes GPU kernels, so the same open model serves faster. On Artificial Analysis for DeepSeek V4 Flash, it's the #1 provider for output speed of all 6 benchmarked providers, about 2x DeepSeek's own endpoint, and the lowest latency too. Live on Infron: DeepSeek V4 Flash and GLM 5.2. Try them: infron.ai/provider/makora Benchmark: artificialanalysis.ai/models…

There’s hope in hard questions.

I keep hearing people argue that continual learning on enterprise usage data will solve the data bottleneck for foundation models. No Fortune 500 will allow this. Their enterprise data is their competitive advantage. Authorizing a foundation model to train on it is like giving a competitor access to your Slack workspace and GitHub repository.

SHOCKING: Many neocloud executives we spoke with feel that if they have non-NVIDIA networking gear on their cluster, or if their cloud has an AMD GPU or TPU offering, NVIDIA will retaliate. They feel that retaliation includes not giving early allocation or no longer supporting a potential IPO/VC raise.(1/3)🧵

I was once pitching in a board room at a top 3 VC firm for a $15M Series A. 12 people in the meeting. One of the GPs fully fell asleep. Out cold for 30+ minutes. Nobody acknowledged it. Everyone just kept going. I kept presenting my Series A slides to an unconscious man in a Herman Miller chair and somehow that was considered normal. That's venture capital. You might fly across the country to perform for people who may or may not be conscious. It's a dance. And sometimes you lead and sometimes you follow and sometimes your partner is unconscious. If you're raising right now, just know: every founder has a story like this. The process is weird. The power dynamic is weird. You're not crazy for thinking it's weird. No one talks about it because they want to continue raising. But I'm happy to stick my neck out there. It is weird.

You just won a 2-week, all-expenses-paid vacation to New York. But there’s a catch: you have to stay within one region the whole time. What are you picking?

ICYMI we showcased our 10x faster than real-time video generation demos at TT-Deploy ft. Tenstorrent Galaxy Blackhole in collaboration with @prodialabs . - 10x faster real-time high-quality video gen - infinite movie w/ real-time prompting

Atlas is open source! An inference engine written from scratch in Rust + CUDA. No PyTorch, no Python, no 200-dependency install dance. <2 min cold start, BUILT for GB10 Qwen3.6-35B at 130 tok/s on a single DGX Spark. Demo powered by @Gradio 🧵 atlasinference.io

We made LLM inference a lot faster. 1/n Speculative decoding throws away draft tokens when they don't match. What if you re-weighted your samples instead using importance sampling? We introduce SMC-SD, sequential Monte Carlo speculative decoding, an approximate sampler that speeds up inference by 2.36× over SOTA speculative decoding, while producing a controllable Pareto frontier of speed and approximation accuracy. ArXiv: arxiv.org/abs/2604.15672 minimal repo release: github.com/abdelfattah-lab/s… Joint work with Mauricio Barba da Costa (@mauricio_b25181), Chi-Chih Chang (@CCCCC1009CCC ), Cameron Freer, Tim Vieira (@xtimv), Ryan Cotterell, and Mohamed Abdelfattah (@mohsaied).