Fetching from the wire…
Public story · 2026-08-16 · high
Checkpoints live as hidden git refs, so a rewind resets code and the AI's conversation in the same step, with no telemetry.
Why now: Waku surfaced in the Aug. 16 coverage after picking up 32 Hacker News points.
Waku wraps the coding agent CLIs you already run and rewinds code and conversation together, per its site. That fixes a real trust problem for developers: roll back a bad diff, and the agent still talks about code that's already gone.
Checkpoints live as hidden git refs, so rewinding the working tree and the provider conversation happens in the same step. Roll back a commit, and Waku resets the AI's memory of that commit too.
It's built natively for macOS on Rust and GPUI, the GPU-accelerated framework behind Zed. The app normalizes sessions, transcripts, tool activity, and checkpoints from different CLIs into one provider-neutral model, so switching agents doesn't mean losing history. Everything stays local: no telemetry, no cloud sync.
It picked up 32 points on Hacker News. Performance claims come from a single source and haven't been verified independently, so treat any speed comparisons with caution.
The interesting claim here isn't the UI. It's that context desync, not model capability, breaks trust in agent coding sessions. If you've rolled back a bad diff and watched the agent reference code that no longer exists, that's the bug Waku targets. Whether a wrapper layer can fix a problem that's really about how CLIs manage state is the open question. Worth watching: whether major CLI vendors build checkpoint-and-rewind natively and make a wrapper like Waku unnecessary.
Each link below shares sources, entities, or timing with this story.
Foundry-Local, alongside community runtimes like LocalAI and cua's sandboxes, points to a first-party push to run agents locally with GPU acceleration and OpenAI-compatible APIs. For cost- and privacy-sensitive loops that fan out lots of cheap subagent calls, a local runtime b...
Warp released its client codebase under AGPL-3.0, surged to 56,000 GitHub stars and #2 on GitHub Trending. But the real story isn't the open-sourcing. It's the repositioning. Warp isn't calling itself a terminal anymore. It's an "agentic development environment." The product n...
Blaizzy/nativ (1,163 stars, Swift, MIT, macOS 26+) comes from the mlx-vlm author and bundles that server into a SwiftUI app that discovers MLX models already in your HF cache. It exposes OpenAI-compatible chat, Responses, image, audio and model endpoints plus Anthropic Message...
5.8k stars, Rust, macOS, ~40 MB versus ~67 MB upstream, keeping full Lua customization. Command-failure recovery with suggested fixes applied via Cmd + Shift + E, natural-language-to-command via # <description>, preconfigured integration for Claude Code, Codex, Gemini CLI, and...
Most coding agents run on JavaScript runtimes and eat 300MB of RAM just sitting idle. Zerostack is built in pure Rust, uses ~8MB on idle, ~12MB while working, and posts 0.0% idle CPU. It hit 1.0 on crates.io this week with 474 points and 252 comments on Hacker News. The archit...
MemPalace (~53.6K stars) claims best-benchmarked, against mem0 (~57.8K) and claude-mem (~80.8K) (GitHub). Agent memory went from experimental nicety to a competitive subcategory with published benchmarks. Pair this with the local-first angle: Mnemo offers a Rust + SQLite + pet...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.