OpenClaw Migrates Web Frontend from Lit to Solid 2.0 Using AI Agents

X postPeter SteinbergerAnnouncement

Peter Steinberger shares how OpenClaw migrated its web frontend from Lit to Solid 2.0 in roughly two weeks using an automated agent loop across approximately 600 PRs with minimal human supervision.

Key points

  1. The migration took about two weeks in total: roughly six days to fully remove Lit and another week to refactor components into idiomatic Solid 2.0.
  2. The refactoring ran mostly automatically via an agent loop generating around 600 PRs, requiring check-ins only every other day.
  3. To de-risk the transition, OpenClaw temporarily shipped an interim release running both frameworks simultaneously.
  4. Steinberger cited Google moving Lit to the Linux Foundation as a signal of soft deprecation and slowed development as the main reason to leave Lit.
  5. Key outcomes include smaller JS and CSS footprints, fine-grained reactive updates, and a faster, more responsive user experience.

Why it matters

This represents a real-world demonstration of autonomous agent loops tackling large-scale architectural codebase migrations, reducing months of tedious framework porting down to days with minimal developer intervention.

Analysis

Beyond the runtime performance gains of Solid 2.0's fine-grained reactivity, this case highlights the evolving role of developers as supervisors or advisors to coding agents. With well-documented libraries, autonomous agents can reliably execute dual-framework migration strategies and refactor idiomatic code across hundreds of incremental PRs.

Original post

Peter Steinberger 🦞

@steipete · 20 October 2026

Two weeks ago I kicked off a loop to move OpenClaw's Lit-based Web frontend to Solid 2.0. Took ~6 days until we could remove Lit, and another week to move patterns to idiomatic Solid. I checked in with the session every other day, overall the process was mostly automatic. Took t.co/yUs40F1C5F

897 likes99 replies322 bookmarks113.2k views

View on X
More posts in the thread (11)
  1. @Infoxicador @RyanCarniato Solid 2.0 is absolutely dope!
  2. @Mike_AscendX Yeah we have more users and downloads than ever. Not everyone wants to give companies control over their data.
  3. @RyanCarniato No idea about diagnostics, I would have to ask my agents. The docs are great tho, almost zero intervention required.
  4. @ian_finlay yeah we even shipped a version that shipped both frameworks to de-risk the transition
  5. @BenEinwechter Google moved it to the Linux Foundation; that’s usually a sign of soft deprecation, and development slowed down.
  6. @Terabytesoftw More like an advisor
  7. @dkatsiros yup! You get a similar experience with the website, but I like having a separate app for it in the dock, and running the app enables computer use, better web previews and other native features.
  8. @giosullutrone99 Yeah the old web UI was not good; that was the focus for the last 3 months
  9. @CrypticTac @Mike_AscendX Yup, you see it right there in the screenshot. Middle part is a widget.
  10. @velismo It was my second choice, but didn’t make the cut, I like Solid’s approach to fine-grained updates better. Performance matters.
  11. @bc1pjordi Click on the link in my post, there’s a full page with reasons why it’s not the best choice for this.
Selected replies (5)
  • @RyanCarniato @steipete I'd love to know more. We've been trying really hard to make Solid 2 agent friendly. Not sure if you ended up using any of our diagnostics in dev etc... But congratulations on the move, and pleased you've seen positive results.
  • @janpfranke @engineerforai @steipete same pain here. what helped me: one shared postgres server, every worktree gets its own db from a template. package download caches are shared too. and a green run only counts for that exact tree, so the full suite never runs twice on the same code
  • @janpfranke @engineerforai @steipete red goes back to the same worker with the failing output, not into the whole loop again. green goes to the reviewer with the results attached, and the reviewer doesn't test again. and yes, unit, e2e and lint, all in that one gate run wrote down the whole setup here: t.co/pB4vJXvKQU
  • @engineerforai @janpfranke @steipete Thanks ill check it out. I run t.co/hFiis47ity . It has very clear routing. The issue is running test suites in parallel. Pretty resource intensive. At least in my project it is. There's intelligent sharing of db test instances, lots of iterations optimizing.
  • @Mike_AscendX @steipete Is openclaw still a thing? Beside HybriqsOS and Grok Bot

Mentioned links

Summary generated by AI (Gemini 3.8 Flash, 21 October 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • X post:Peter Steinberger

    OpenClaw Deletes 400k Lines of Tests Using Dedicated Agent Skill

    Peter Steinberger reported that OpenClaw pruned around 400,000 lines of its own test code with minimal impact on code coverage. The cleanup was driven by an agent skill called 'test-audit', which targets low-value, redundant, or implementation-coupled tests.

    10.8kbookmarks1.3mviews

    AI & LLMs· News

  • Article:Hannes Rudolph

    OpenClaw 2.0: Major Update Introduces Simplified Setup and Shared Sessions

    Open-source project OpenClaw has announced OpenClaw 2.0, representing the largest release in its history. Developed over nearly seven weeks by 933 contributors, the update incorporates more than 16,000 pull requests. What started as an effort to streamline installation and rebuild the browser interface turned into a comprehensive overhaul of the platform, including new multiplayer collaboration features.

    AI & LLMs· Announcement

  • Article:Fred K. Schott

    Flue 2.0: Agent Framework Adopts Hooks Architecture

    Flue 2.0 rebuilds the TypeScript agent framework around a React-inspired hooks API. This allows developers to build dynamic agents that can adapt models, tools, and persistent state at runtime as conversations evolve.

    AI & LLMs· Announcement

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.