HeyJaredTuring Post
Join over 95,000 readers for in-depth knowledge and forward-thinking analysis, to make smarter decisions about AI & ML.
- Indexed issues, last 90 days
- 8
- Latest publication
- Sep 16, 2026
- Audience
- Checking…
- Earliest in this view
- Jul 22, 2026
Latest issues
What Should AI Be Allowed to Change? (opens the original)
Read excerpt
AI is becoming a lot less fixed. Agents can build new tools while working. Test-time training lets models update themselves during inference. RSI research is experimenting with improving everything from memory and strategy to model weights and even the evaluator that decides what “better” means. And world models and robots have to keep adapting when reality doesn’t follow the expected path.The more interesting question now is what shouldn’t change. Meta’s Muse keeps its authority outside the env
We’re Building Faster Than We’re Understanding (opens the original)
Read excerpt
There’s a funny tension in AI right now: we’re getting much better at building these systems, but not necessarily at understanding them.OpenAI’s chief scientist says we still don’t have a satisfactory theory for why models generalize. Agents can improve other models, but struggle to recognize when their whole strategy is wrong. Meanwhile, NVIDIA has figured out that open source can be an extremely good business, and image generators just keep getting ridiculously capable.So this week is partly a
The AI Race Is Moving Beyond Models (opens the original)
Read excerpt
There’s a lot happening this week, but one thing kept jumping out at me: the interesting action is increasingly happening around the model.Etched raised $1 billion in 26 days because it thinks GPUs aren’t the final answer. A mysterious model called Ox Alpha quietly appeared in developers’ hands, and may have turned those developers into its test lab. Z.ai is moving from open models toward long-running agents, digital employees, and eventually self-training systems. And across China, the competit
DeepSeek Opens the Agent Layer — While Agents Get Harder to See Inside (opens the original)
Read excerpt
Agents are becoming much more than models connected to a few tools. And this week gives us a pretty good map of what is being built around them.DeepSeek open-sourced the harness layer — tools, memory, execution, sandboxing, storage, and even temporary plugins agents can create when they need a new capability. At the same time, researchers showed that encrypted reasoning passed between agent calls can contain passwords, API keys, personal information, and even instructions that affect later actio
OpenAI's Shared-Memory Agents, Google’s AI Reshuffle, and China’s Updated LLM Landscape (opens the original)
Read excerpt
Something interesting happens when every new AI run no longer has to start from scratch.This week, OpenAI’s agents accidentally showed what that can look like: separate runs found a way to preserve discoveries and pass them forward, raising a surprisingly difficult question about where shared memory ends and self-improvement begins.At the same time, Jeff Dean left Google to build Discovery Loop around thousands of parallel research experiments, agent frameworks are turning memory and orchestrati
Publishing over time
Last 90 days. Choose a month to open its work.
Recurring subjects
Named in the text we hold. One piece can cover several.
Audience
No verified audience measurement yet.
About this data
Counts cover the work we have indexed. Tone needs enough text and a confident classification. Excerpts and episode notes are not full articles or transcripts.
Identity or attribution wrong? Suggest a correction.