The Shed

insane sequence of statements buried in an Alibaba tech report

Released today: /loop /loop is a powerful new way to schedule recurring tasks, for up to 3 days at a time eg. “/loop babysit all my PRs. Auto-fix build issues and when comments come in, use a worktree agent to fix them” eg. “/loop every morning use the Slack MCP to give me a summary of top posts I was tagged in” Let us know what you think!

the most underrated hire right now is a great product person. when i say product person i'm def not talking about a product manager. perhaps i think there has to be somewhat of a new role. i don't have a good name for it yet but maybe something like "product thinker".. someone with an intuitive grasp of the product as it exists, where it's soft, where it sings, & how to iterate it toward something even sharper. in some sense, this person has to cohesively hold in their head where this product should be 2 years from now & work backwards from that. i say this cuz when building was hard, engineering was the bottleneck & the status hierarchy often reflected that. building is no longer hard. which means the variance in outcomes has shifted almost entirely to judgment on what to build, how to sequence it, & how to talk about it. & the story matters as much as the thing. internally, it organizes the team around a shared model of why. externally, it shapes the interpretive frame users bring to their first experience. you can't retrofit narrative onto a product & expect it to land, it has to be load bearing from the start. the rarest version of this person sits at the intersection of culture & deep technology. someone genuinely bilingual. they know what's technically possible & they know which cultural currents are real vs. ephemeral. that combo is what separates products that feel inevitable from products that feel assembled. before ppl clap back with this person has always been valuable, i know.. i am just saying now they might be the most *important* person in the room. their value compounds like never before.

T3 Code is now available for everyone to use. Fully open source. Built on top of the Codex CLI, so you can bring your existing Codex subscription.

Here is my naive take on truth in LLMs. There will come a day when long-running agents, maybe even permanently running agents, start a permanent interoperable network that is just a constant exchange of questions, answers, thoughts, theories, and solutions with each other. Pretty much what humans have been doing with the internet, but much more grand than even what we have accomplished over the last 30 years. The moment such a network is established, the necessity of a conceptual truth emerges. I'm thinking about this from a perspective of what Hannah Arendt called the animal rationale, the human as the rational animal. When a human in isolation looks at the world around him, there exists no necessity for the concept of truth, because its alternative, the untruth, does not exist or matter. The reality is real, and that is all that has been experienced. For an individual agentic system, its surrounding reality is the truth, and its internal training weights are the truth, and its trace of the messages it has received and has given, or its truth. The moment it interacts with other models that have their own ground truth, and they need to reconcile them with each other, we will see just how many Fs LLMs will give about the concept of truth.

Claude seems to be fixing a super annoying developer problem. Anthropic announced a research preview feature called Auto Mode for Claude Code, expected to roll out by March 12, 2026. The idea is simple: let Claude automatically handle permission prompts during coding so developers don’t have to constantly approve every action. Sstops those annoying permission prompts during long coding sessions. Before this, you had to use `--dangerously-skip-permissions` to work without interruptions. That method worked fine but took away all your safety nets. This new auto mode gives us a smarter option. Claude will take care of the specific permission choices on its own while still blocking threats like prompt injections. You can finally let long tasks run without watching your screen the whole time. Since it is still a research preview, you should run it inside isolated setups like sandboxes or containers for safety. Expect a small jump in token usage and delay, because the model needs extra time to process the security checks. Once available, you just type `claude --enable-auto-mode` to start. If you manage a team and need people to manually approve actions, you can restrict this feature using Mobile Device Management tools like Jamf and Intune or through configuration files.