Best new AI, design and productivity resources: day 34

Yesterday's additions cover three things you might care about: organizing work, seeing what AI agents actually do, and building better interfaces. Treat them as curated starting points, not endorsements of every claim the projects make. Follow the original links and see whether each one fits how you work.

Fresh ideas, and projects that keep moving

Prepping a product demo and stuck on a blank page? What Ships collects startup launch videos from X, each linking back to the original post. Browse a handful to see how other teams introduce their products. https://whatships.com/

TaskView is a self-hosted project and task management option for teams that want direct control over their workspace. It makes more sense for ongoing work than for one-off personal to-dos. https://github.com/Gimanh/taskview-community

Mulmoclaude is a local-first AI assistant that keeps memories and small personal apps in files on your own computer. If owning your working context matters more to you than having a fully hosted assistant, give it a look. https://github.com/receptron/mulmoclaude

Seeing what your AI agents actually did

Flyto Core is built around recorded, replayable browser automation. That means you can inspect evidence of what an agent did, instead of relying only on its final message. Try any automation on low-risk work before you trust it with consequential tasks. https://github.com/flytohub/flyto-core

Building with Django? Django Orbit adds agent-readable traces and debugging to Django applications. It's a focused pick for teams already on Django, not a general-purpose monitoring tool. https://github.com/astro-stack/django-orbit

Threadport packages portable, verifiable work state for coding agents, and Agentlas OS organizes specialist agents around tasks. Put simply, Threadport helps with continuity between sessions, while Agentlas OS is about coordinating multiple agents. https://github.com/Frankie-Xu/threadport

Agentlas OS lives here: https://github.com/agentlas-ai/Agentlas-OS

Railtracks and Dive both help you build agent systems, just in different ways. Railtracks is a framework for resilient agent workflows, and Dive is for writing agents in Go. Pick based on your stack, then test the behavior in your own environment. https://github.com/RailtownAI/railtracks

Find Dive at https://github.com/deepnoodle-ai/dive

Building interfaces that stay consistent

Generated interfaces commonly go inconsistent in two places: structure and movement. Balsa UI tackles structure, with design systems AI agents can read, from tokens to components. Springs Studio handles movement, with agent-readable easing and motion tokens. https://balsa-ui.com/

Springs Studio is at https://www.springs.studio/

Meshfont is a free, browser-based playground for designing fonts. Best Free Fonts, meanwhile, is a collection for finding typefaces that already exist. Want to experiment with letterforms? Use Meshfont. Need to choose a font quickly? Use Best Free Fonts. https://meshfont.com/

Best Free Fonts lives at https://bestfreefonts.com/

ForgeUI and 21st offer interface components and templates, and Paper gives you a place to design and share work. Before a component goes into a production project, check its code, licensing and compatibility. https://forgeui.in/

21st is at https://21st.dev/

And Paper is at https://paper.design/

Want more motion? Theatre.js gives you an animation editor, Unicorn Studio offers interactive graphics, and Spline handles collaborative 3D design in the browser. They each play a different role, so pick the one that matches the medium you're working in. https://www.theatrejs.com/

Unicorn Studio lives at https://www.unicorn.studio/

And Spline is at https://spline.design/

A few more focused additions

CoSAI Risk Map gives you a framework for thinking through the security risks of an AI system. Agent Rules Skill helps generate repository instructions for coding agents, and AI-First Toolkit collects skills for Claude Code and Codex. Before any of them get access to your code or accounts, review the generated instructions and third-party skills. https://github.com/cosai-oasis/secure-ai-tooling

Agent Rules Skill is at https://github.com/netresearch/agent-rules-skill

And AI-First Toolkit lives at https://github.com/techwolf-ai/ai-first-toolkit

There's more, too: LiveContext CE for self-hosted AI workflows, People Context for local-first personal memory, OpenExecutive for virtual executive-team workflows, and Kiln for agent-assisted 3D printing. You'll find all of yesterday's additions, with their original links, in the complete directory. https://bestproductivityresources.com/directory