rUv ruvnet

RuView exceeds MultiFormer on MM-Fi (WiFi-CSI pose) — 81.63% torso-PCK@20

Controlled, protocol- and metric-matched claim:

RuView's CSI-Transformer achieves 81.63% torso-PCK@20 on the MM-Fi random_split WiFi-CSI 2D pose benchmark, exceeding MultiFormer (72.25%) and CSI2Pose (68.41%) on the same protocol and same metric. Absolute gain +9.38, relative +13.0%.

⚠️ What this is NOT. This is a protocol-matched MM-Fi random-split result, not solved real-world generalization. Random split contains temporal/subject-adjacency effects common to this benchmark family. Cross-subject (official split, shared rooms) is ~64%; the hard frontier is cross-ENVIRONMENT (new room) at ~11.6%; cross-subject pose is the real deployment frontier. We do not claim universal WiFi-pose SOTA (e.g. WiFlow reports 97% PCK@20 but on a separate 5-subject self-collected set — not an MM-Fi leaderboard comparison).

GAIA L1 iter 65 — Co-Sight DAG: 3 tool fixes + full 53Q validation

Branch: feat/iter-64-cosight-dag | PR: #2218 | Issue: #2156

Score: 25/53 (47.2%) — cost $1.16 (3.9× cheaper than single-agent)

Benchmark progression

iter	Architecture	Score	Cost	Notes

Ruflo Agent Capability Benchmark — Detailed Overview

Companion gist for PR #2163 and the Dream Cycle 2026-05-27 capabilities-scan finding (#2156).

Session date: 2026-05-27 · Commits landed: a6dd4ab3d, dede70efd, 88743c482, 7e3ec89e4, a7dfdec4c · Branch: feat/2156-agent-benchmark-suite

TL;DR

	Before	After

ruflo v3.8.0 — SOTA agent-framework benchmarks

How fast is the framework layer of an AI agent system? We benchmarked ruflo 3.8.0 against LangGraph 1.2.1, AutoGen 0.4.9, CrewAI 0.80.0 on an identical workload, on two operating systems, with a stub LLM (so we measure framework overhead, not model latency).

The short version: ruflo is faster than the comparators on cold start, single-turn dispatch, and memory footprint by 1.3× to 1,953× — on both macOS and Linux. On the two dimensions where CrewAI shows a slight edge (compose_50_tools and N=10 parallel), CrewAI's numbers are proxied lower bounds — its real dispatch requires an LLM call that adds seconds.

TL;DR — who wins each dimension

Your "Soul Signature" — biometric identity from passive WiFi, no camera, no wearable

Project: RuView / WiFi-DensePose · Date: 2026-05-24 · Spec: docs/research/soul/ in the repo

The pitch in one paragraph

Your body is constantly bouncing radio waves around the room you're in. WiFi routers fill that room with 2.4 / 5 / 6 GHz signals; you reflect, absorb, and re-scatter them in a pattern that's measurably unique to you — your heartbeat, your breathing, your posture, your gait, your skeletal proportions, the way your chest cavity flexes when you inhale. RuView captures those reflections from a $9 ESP32 sensor and fuses them into what we call a Soul Signature — a single per-person electromagnetic fingerprint. No camera. No microphone. No fingerprint reader. No wearable on your wrist or in your ear. The system just notices you in the room and knows who you are — through walls, in the dark, while you sleep.

@claude-flow/guidance — performance benchmarks (rigorous baseline + 4 iterations to SOTA)

Package: @claude-flow/guidance@3.0.0-alpha.3 · 15+ source files, 1,331 tests Repo: https://github.com/ruvnet/ruflo · branch perf/guidance-phase-1-hotpath-optimizations · PR #2103 Date: 2026-05-22 · Node v22.22.1 on darwin-arm64 Methodology: 5-trial median, 50-2000 iterations per trial depending on N, warmup phase to trigger V8 JIT tier-up

TL;DR — M4 quantization delivers a 2.70x end-to-end speedup at N=1000

RuFlo Graph Intelligence Engine — `ruflo-graph-intelligence@0.1.0-alpha.1`

Real-time relationship intelligence with complexity-aware execution. Built on sublinear-time-solver@1.7.0. 12 graph wedges, signed reasoning artifacts, federation-distributable PageRank vectors. 104/104 tests. MIT.

Introduction

Most AI memory systems recompute everything. RuFlo computes only what changed enough to matter — and only at the depth the runtime can afford.

ADR-123 — RuFlo Graph Intelligence Engine: real-time relationship intelligence with complexity-aware execution

Status: Proposed (2026-05-18) — revised 2026-05-19 to track upstream sublinear-time-solver@1.7.0 Date: 2026-05-18 Authors: claude (drafted with rUv) Related: sublinear-time-solver@1.7.0 (crates.io sublinear@0.3.0, github, 1.6.0 announcement gist, upstream sublinear ADR-001 "Complexity as Architecture"), eleven-wedge research gist, ADR-103 (witness temporal history), ADR-104 (federation wire transport), ADR-105 (federation state snapshot), ADR-118 (AIDefence 2.3.0), ADR-121 (embeddings RuVector u

🚀 sublinear-time-solver 1.6.0 — security fix, correctness overhaul, full CI

Released 2026-05-18. Tags: v1.6.0. Package: sublinear-time-solver (npm), consciousness-explorer (npm), sublinear (crates.io).

TL;DR — should you upgrade?

Yes, immediately, if any of these apply to you:

You expose the MCP tools to anything other than trusted local clients. This release closes a remotely-exploitable arbitrary file-write (CWE-73) reported by @BruceJqs in issue #19. The same bug existed in two MCP tools — export_state (consciousness-explorer) and saveVectorToFile (main package). Both are fixed in 1.6.0.

	{
	"n_features": 6063,
	"train": 4850,
	"heldout": 1213,
	"heldout_triplet_acc": {
	"raw_features_baseline": 66.42,
	"random_encoder_baseline": 69.56,
	"trained_encoder": 82.33
	},
	"int4_encoder_bytes": 4640,