It's been *almost* a bit quiet around LLM architecture releases in the past two weeks Interesting tidbit is the parallel block design. Via the Cmd-A the tech report "equivalent performance but significant improvement in throughput compared to the vanilla transformer block."
AI
-
Anthropic gets up to GB200 capacity at Colossus 2 with SpaceX
By
–
Anthropic SpaceX Anthropic is getting up to GB200 of capacity in Colossus 2 in June as a part of the expanded agreement with SpaceX. Partnerships are huge unlocks
-
Railway’s agent-native cloud for AI coding agents
By
–
🆕Railway's Agent-Native Cloud: 3M users, 100K signups/week, $200K+ coding agent spend, production forks, & the death of PRs https://t.co/kB3yqOWQm1@Railway founder @JustJake explains why AI agents need a new cloud, how Railway moved the vast majority of workloads onto its own… pic.twitter.com/WYa8RHxPTM
— Latent.Space (@latentspacepod) 20 mai 2026Railway's Agent-Native Cloud: 3M users, 100K signups/week, $200K+ coding agent spend, production forks, & the death of PRs https://
latent.space/p/railway @Railway founder @JustJake explains why AI agents need a new cloud, how Railway moved the vast majority of workloads onto its own -
SpaceX Expands AI Compute-as-a-Service Offering
By
–
As the recently expanded partnership with @AnthropicAI demonstrates, @SpaceX is offering AI compute as a service at significant scale. We are in discussions with other companies to do the same. Over time, especially with orbital data centers, we expect to serve AI at extremely
-
Dynamic Fine-Tuning (DFT) combines SFT loss with forward KL penalty
By
–
This is so neat! Dynamic Fine-Tuning (DFT) reweights the SFT loss by the model's own token probability, which creates a feedback loop. So they added forward KL to penalize any token the base finds likely, but the policy has pushed toward zero probability. DFT and forward KL
-
Inquiry into AI model training methodology and data augmentation
By
–
i would be pretty interested to know if anything other than scale had changed and want to know about the training, whether there was a lot of symbolically generated augmented data, etc
-
Abacus AI ChatLLM Integrates RouteLLM for Model Comparison
By
–
RouteLLM – Use The Best AI For Your Prompt Opus 4.7, GPT 5.5 And Gemini doing their best work!! All running inside Abacus AI ChatLLM – one workspace to access, compare, and switch between the top frontier models without juggling tools. Opus 4.7 → deep reasoning, long-form
-
AI increases information quantity but decreases quality
By
–
Great, AI has increased the quantity of information and decreased the quality. Just what we needed.
-
Benefits of Lightweight Code Interpreter Environments for AI Development
By
–
code interpreter is a light weight code execution environment lets you do:
– RLMs
– programmatic tool calling
– more! without having to spin up a full sandbox we'll be writing a lot more about the use cases here, but check it out! -
Anthropic Projected to Double Revenue to $10.9 Billion
By
–
Even bigger win for Anthropic: Anthropic’s revenue is set to more than double to $10.9 billion in the second quarter, an explosive rate of growth that will help it turn an operating profit for the first time. About $500m in profit. Nuts.