Ordewell is a tool that converts a single goal into an ordered plan of coding-agent tasks, each with configurable runner, model, and mode, then executes and verifies them. It provides a transparent, editable plan before execution, supports multiple AI providers and runners, and includes both a CLI TUI and VS Code extension.
A talk examining collaborative planning with software agents, drawing parallels from anthropologist Thomas Gladwin's study of Chuukese navigators who improvise in real-time versus European navigators who plan rigidly in advance. The speaker, a researcher at GitHub Next, argues that plans are weak resources for real-world action and advocates for better tools enabling humans and agents to plan together collaboratively, moving beyond current isolated CLI-based question-answer interfaces.
The Awesome Calendar is a free printable linear calendar that visualizes the entire year as a continuous 1.5-metre timeline. Users download a PDF, print it on A4 sheets, cut the strips, and join them together to create a single timeline showing past and future as points on an axis.
LifeTrajectory is a free longevity assessment tool that provides personal orientation through a 12-question questionnaire. Created by Andreas Holzhäuser, it offers early insights into life trajectory and planning.
TattooPreview AI is a planning workflow tool that uses AI to generate three distinct tattoo concept directions from a written description, helping clients prepare clearer briefs for tattoo artist consultations. The platform guides users through meaning, mood, placement, and linework stages without claiming to produce finished tattoos, requiring professional artist review and redraw.
A design document articulates hard problems and coordinates decisions among development teams before implementation begins. Writing a design doc is most valuable for complex, risky, or multi-team projects, with investment level depending on project scope, risks, and team culture. Key decisions for design docs involve assessing the cost of being wrong—permanent architectural choices warrant detailed specification, while trivial implementation details do not.
A design document is a planning tool that helps teams coordinate work and avoid costly implementation mistakes by articulating hard problems and key decisions before coding begins. The investment in a design doc should match project complexity, risk, and team coordination needs, with the key question being whether getting a decision wrong would be expensive to fix.
Subnet Calc is a visual subnet planning tool for teams that helps users calculate and manage IP address subnets. The application stores designs locally in the browser and offers features like IP address input, prefix length configuration, and project saving.
AI has removed the friction that once forced deliberate planning, creating a choice between two approaches: building structure first before delegating to AI (improving efficiency by ~25%), or skipping planning and generating output immediately (reducing efficiency by ~18% due to verification costs). The critical skill is now planning and structuring work before delegation, not using AI itself.
Researchers from Google, Georgia Tech, and Peking University introduce the Procedural Graph, a framework that organizes procedural knowledge into (procedure, relation, procedure) triplets to guide LLM agents through long-horizon planning tasks. The graph self-evolves by contrasting failed and successful trajectories, automatically refining its structure to improve agent performance without manual engineering across multiple datasets and task types.
Procedural Graphs organize procedural knowledge for LLM agents through (procedure, relation, procedure) triplets, guiding agent actions at each decision step. The framework is self-evolving: an LLM refiner learns from failed and successful trajectories to edit the graph's topology, building effective execution structures from minimal skeletons that match or exceed hand-designed alternatives.
Danijar Hafner, a 31-year-old AI researcher, is launching a stealth startup in San Francisco to develop robots that can handle unexpected real-world situations through model-based reinforcement learning. His approach trains AI agents in simulated world models to plan ahead and navigate unfamiliar environments without extensive real-world trial-and-error, advancing the goal of deploying humanoid robots in human spaces.