Who's Who of AI

Chris Paxton

282 trust commentator @cpaxton.bsky.social · 6,773 followers
Writing about robots https://itcanthink.substack.com/ RoboPapers podcast https://robopapers.substack.com/ All opinions my own
What they're sharing

Articles & links

↻ Chris Paxton reposted
Max Woolf @minimaxir.bsky.social

son of a bitch blog.google/innovation-a...

DiffusionGemma: 4x faster text generation blog.google
AI Weekly's analysis →
  • DiffusionGemma generates 256 tokens per forward pass using bidirectional attention, reaching 1,000+ tokens/sec on a single H100 GPU.
  • With only 3.8B active parameters during inference and an 18GB VRAM footprint when quantized, it runs on consumer hardware without server-grade resources.
  • Google recommends DiffusionGemma only for speed-critical workloads like in-line editing and code infilling, not for applications requiring maximum quality.
Read full analysis →
 View on Bluesky →
↻ Chris Paxton reposted
@isolyth.dev

holy shit, Anthropic is about to become profitable. Not profitable minus training costs, just profitable, period. Today is not a good day for denialists at all.

wsj.com
AI Weekly's analysis →
  • Anthropic projects $10.9B in Q2 2026 revenue, more than doubling Q1's $4.8B in a single quarter.
  • Claude Code surpassed $1B in annualized revenue within six months of its launch, driven by enterprise developers.
  • An expected $559M operating profit marks the first quarter Anthropic covers costs without relying solely on outside capital.
Read full analysis →
 View on Bluesky →
↻ Chris Paxton reposted
adinayakup.bsky.social @adinayakup.bsky.social

GLM 5.2 is here 🔥 huggingface.co/collections/... ✨ 753B - 1M context ✨ MIT license ✨ GLM IndexShare: reuses the indexer across layers, 2.9x fewer FLOPs/token at 1M ✨ AIME 2026: 99.2 (beats GPT-5.5, Claude Opus 4.8) ✨ vLLM / SGLang / Transformers supported

GLM-5.2 - a zai-org Collection huggingface.co
AI Weekly's analysis →
  • Z.ai released GLM-5.2 on Hugging Face, a 753B-parameter open weights model under an MIT license with a 1M-token context window.
  • An IndexShare design reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9x at 1M context length.
  • Self-reported scores include 62.1 on SWE-bench Pro, 82.7 on Terminal Bench 2.1, 99.2 on AIME 2026, and 91.2 on GPQA-Diamond.
Read full analysis →
 View on Bluesky →
↻ Chris Paxton reposted
Sung Kim @sungkim.bsky.social

New SWE benchmark, called DeepSWE. I'm curious why Cursor is not included. Anyway, I stand by my statement that Claude Code (Claude Opus 4.7) is better at first 80% and Codex (GPT 5.5) is better at remaining 1,000%, at this moment. deepswe.datacurve.ai

DeepSWE deepswe.datacurve.ai View on Bluesky →

We are finally seeing the launch of cheap, capable robot hardware that is at consumer price levels. new announcements from Weave, Nori Robotics, and Bracketbot show the wave of future consumer robot products that will live in your home and do your chores. more: open.substack.c…

Rise of the Cheap Robots itcanthink.substack.com
View on Bluesky · ♥ 61 ↻ 8 ↩ 5 · 6d ago
Their own posts

Recent commentary

The best AI systems being closed off and restricted to just a handful of elites, is basically my worst case AI doom scenario. I really hope the z ai guys werent kidding about open source fable/mythos soon

View on Bluesky · ♥ 141 ↻ 17 ↩ 8 · 13d ago

AI is good at cutting edge math, scientific research, and software engineering, but still cant install a new drain in your sink. This is because, presumably, being a plumber requires a human soul

View on Bluesky · ♥ 98 ↻ 4 ↩ 8 · 28d ago

The us government is picking customers who can access gpt 5.6

View on Bluesky · ♥ 83 ↻ 11 ↩ 5 · 13d ago

Post training deepseek with a 1k gpu Huawei cluster -- no nvidia -- feels like a serious milestone to me "Intelligence too cheap to meter" cant arrive if there's an nvidia/tsmc bottleneck, even if its better for shareholders

View on Bluesky · ♥ 50 ↻ 3 ↩ 2 · 27d ago

Eno -- the new humanoid robot from Genesis AI

View on Bluesky · ♥ 27 ↻ 1 ↩ 4 · 22d ago

Ai is a scam that will never be profitable

View on Bluesky · ♥ 25 ↻ 1 ↩ 0 · 49d ago

I think an AI pause is a bad idea. Isn't mythos their most aligned model ever? This feels like it will be self correcting, if RSI hits itll have better ideas for handling RSI than we do

View on Bluesky · ♥ 16 ↻ 0 ↩ 5 · 32d ago

Its being reported that the iran war reparations are all SpaceX and OpenAI stock

View on Bluesky · ♥ 17 ↻ 1 ↩ 1 · 20d ago

The US government is forcing anthropic to block all foreign nationals from using Mythos and Fable. This will be a big boost for national and transnational AI programs outside the US, and I fear it will provide further fuel to decoupling.

View on Bluesky · ♥ 4 ↻ 0 ↩ 1 · 25d ago

I generate a lot of AI art for blog post hero images. I can't tell if the model is actually getting worse (almost certainly not) or if I am just getting sick of the style but I FEEL as if the results are worse. Perhaps I need to switch generators or just sketch things myself

View on Bluesky · ♥ 2 ↻ 0 ↩ 1 · 32d ago
Their network

In Chris Paxton's orbit

Center = Chris Paxton. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.