Can we just maybe get "line go up now but might not forever?"
Who's Who of AI
Isaiah Bishop
Why they matter
Tracked through public AI activity and peer connections inside the directory.
- AI signals
- 4 past 30d
- Sources
- 4 distinct domains
- Discussions
- 108 past 30d
- Latest signal
- 5d ago
ML Eng. and econometrics. Honorary Commodities Trader**. more left-posting than normal.
**Regrettably degen trading on here, im sorry
Views dont reflect my employer
What they're sharing
When AI builds itself anthropic.com
Claude Fable 5 and Claude Mythos 5 anthropic.com
View on Bluesky →
Generative artificial intelligence creates delicious, sustainable, and nutritious burgers | npj Science of Food nature.com
nytimes.com
View on Bluesky →
openai.com
Emergent Capabilities Arise Randomly from Learning Sparse Attention Patterns arxiv.org
wsj.com
View on Bluesky →
What Investment Data Implies about the AI Transition nber.org
View on Bluesky →
z.ai
View on Bluesky →
Prime Minister Carney launches AI for All: Canada’s new national artificial intelligence strategy pm.gc.ca
Misalignment Between Backpropagation and the Hierarchy of Brain Responses to Images arxiv.org
View on Bluesky →
Articles & links
Isaiah Bishop reposted
www.anthropic.com/news/claude-...
Isaiah Bishop reposted
Finally, AI finds its ultimate use case. A diffusion model trained on burger recipes "discovers the classic Big Mac without explicit supervision and generates novel burgers optimized for deliciousness, sustainability, or nutrition." ASI= automated slider intelligence www.natur…
AI Weekly's analysis
→
Read full analysis →
View on Bluesky →
Isaiah Bishop reposted
It's interesting. Don't have an entirely clear picture from just this, but I would have questions about the counterfactuals here that I doubt could be answered (would these groups' tactics be substantively different without this access? etc.). www.nytimes.com/2026/07/10/u...
Man look at the change in the sales pitch openai.com/index/gpt-5-6/
Isaiah Bishop reposted
I'm so happy when other people write papers on nondeterministic factors in training. embrace the chaos
AI Weekly's analysis
→
- Emergent capabilities arise stochastically: the same model can gain or fail to gain a capability depending on its random initialization.
- Researchers used Pythia models from 14M to 410M parameters to show attention pattern learning is the key bottleneck to capability emergence.
- More attention heads improved learning efficiency; MLP-Mixer outperformed standard transformers on tasks with complex attention patterns.
Read full analysis →
View on Bluesky →
Isaiah Bishop reposted
“.. what does it say when even Masa thinks it’s too far out there?” @wsj.com $SPCX www.wsj.com/tech/why-one...
Isaiah Bishop reposted
According to a new NBER paper, to justify the AI investment surge we're seeing, the implied productivity gains need to be enormous. The authors calibrate a model where AI-sector productivity rises by roughly 2.7x. Link: www.nber.org/papers/w35290
Isaiah Bishop reposted
GLM-5.2 weights are now out! 1M context, and beats or matches GPT-5.5 and Opus 4.7 on many tasks... I hope to see a comparison between this model and Kimi-k2.7 soon z.ai/blog/glm-5.2 It also contains agentic slime (?)
some truly awful post secondary ideas. and i dont know what theyre thinking with the supercomputer
Isaiah Bishop reposted
👀 arxiv.org/abs/2605.28693
Isaiah Bishop reposted
Their own posts
Recent commentary
Have openAI and Anthropic cracked continual learning? Well its simple, if its what you expect they wouldn't need to keep raising money
Was listening to Dwarkesh but im always struck by this kind of belief that AI will be able to, in his words, start a business and be great at it or "be as good at texas politics as LBJ" and like it just seems to me we are asking too much of generalization
it does kind of feel like a lot of things have become correlated in ways that are undesirable. Idk how to say this but like it was useful when bonds and stocks werent correlated. Useful when US and ex-US werent just different flavours of the AI trade. And like same politically.
Finding myself kind of bored with AI and stocks now. Neither really have the juice atm
The openAI financials, are honestly better than i thought
One of the under-appreciated aspect of LLM debugging is how much progress it makes on things i wouldnt even know to check. Like i do ML, i know python/R, i can mess around with devops stuff but its not why you hire me Claude immediately checked some obscure datadog logs, found an S3 role issue
Agentic as an adjective is a massive ick for me
waiting desperately for stocks, macro, or AI to be like super interesting instead of boring
Could you make a more efficient high level language that only an LLM could understand?
do we think i could get some VC money to teach an AI how to do machining
Their network
In Isaiah Bishop's orbit
Center = Isaiah Bishop. 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.