Fetching from the wire…
Public story · 2026-07-16 · high
It tracks which subagent actually read each artifact and pushes revisions or revocations to every token holder, something a shared file path can't do.
Why now: Waggle went from a July 8 post to 762 stars and a v0.5.3 release by July 14.
Waggle turns pasted context into a 30-byte token that resolves in 39 nanoseconds, per the project's GitHub repo. Multi-agent setups re-paste the same artifact into every subagent's context, whether the reader needs it all or not. Waggle's token stays about 30 bytes no matter how large the artifact underneath is.
Each artifact resolves into a consumer-specific projection sized to the reading model. Only the requested slice travels, under a byte budget; nothing auto-expands into context outright.
The tool, published as modiqo/waggle, also tracks which subagent actually opened its input. It pushes revisions, or revocations, to every token holder when the source changes, something a raw filesystem path can't do.
It runs as an MCP server, a one-line config addition in Claude Code, Codex, or Cursor. The repo lists 39ns for a cache-hit resolve, 39µs for a durable append, and 1.2ms p50 at the edge. It's dual-licensed under Apache-2.0 and MIT, shipped v0.5.3 on July 14, and picked up 762 stars since posting on July 8.
Every agent framework still passing raw file paths between subagents is already behind this, if the revision propagation holds up outside a demo. The repo doesn't say what happens to the byte budget when several subagents request different slices of the same artifact at once. It also doesn't say how a token holder catches up after being offline through a revision.
Each link below shares sources, entities, or timing with this story.
v0.10.0 (~84.8k stars, Apache-2.0) ships no agent of its own and drives whichever CLI you already have, Claude Code, Codex, Cursor, Copilot, OpenClaw, Gemini, Kimi, Qwen, Cline, plus BYOK OpenAI-compatible endpoints, via od mcp install <agent>. It produces single-page HTML pro...
diegosouzapw/OmniRoute added 1,343 stars on July 20, a single MIT-licensed gateway across 268+ providers (50+ free) and 500+ models including Claude, GPT, Gemini, Kimi K3, GLM and DeepSeek, wired for Claude Code, Codex, Cursor, Cline and Copilot. Quota-aware automatic fallback...
A local-first observability dashboard that tails agent session files to show plans, tool calls, file changes and progress in real time across three different coding agents, MIT-licensed, still being pushed today (GitHub). It's the only repo carrying the agent-observability top...
affaan-m/ECC (36.3k forks, MIT) bundles 67 agents, 284 skills, 94 legacy command shims, and "instincts", patterns learned from prior sessions with confidence scores that auto-recall when relevant, plus a .ecc/memory/ markdown vault that's explicitly cross-harness, so context s...
The GTM Co-Founder took #1 with 162 upvotes: a free bundle of Agent Skills that interviews a founder for about 15 minutes, learns the product and market, then walks through a prioritized GTM roadmap covering positioning, first 50 users, launch and pricing. Runs inside Claude C...
QM went up under MIT license. Created July 29. As of the GitHub API check: 8,420 stars, 887 forks. Five days. YC uses it internally across accounting, legal, events, and engineering, including to build QM itself. Every employee and every Slack room gets its own scoped memory,...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.