NGI LogoNGI.video
← Back to all apps

Velocicode

In active development

Point it at slow Python code and it builds faster native versions, keeping only the ones whose output matches exactly.

Velocicode showing its Python hotspot profiler native compiler comparison and parity gate

Current status

Early Python-to-native optimization prototype; its hotspot and parity checks are integrated into the Video Studio engine.

Overview

Automated Python optimization engine that pinpoints the slowest functions in a codebase, compiles them into native machine extensions, and only keeps builds whose output matches the original exactly.

Why it was built

Rewriting slow Python loops in C or Rust by hand takes days of debugging, and unverified auto-generated native code can silently change program results or even run slower.

Why it helps

Profiles real program bottlenecks automatically, tests candidate native builds in an isolated sandbox, and guarantees every shipped `.so` extension produces identical outputs at measurably higher speed.

What you can do

  • Automated bottleneck profiler that focuses compilation effort where it actually saves wall-clock time
  • Multi-compiler build harness that compares native C and Rust extension candidates automatically
  • Strict output-parity verification gate that rolls back any build that differs from Python behavior
  • Zero-downtime hot-reload and AI coding assistant integration for continuous optimization
Velocicode • Core Technology & Feature Architecture

Amdahl Hotspot Profiler, Runtime Trace Harvester & Native .so Parity Swapper

Pinpoints real Python bottlenecks via sys.monitoring, harvests live golden I/O traces, compiles native C/Rust extensions in an isolated sandbox, and hot-swaps only builds passing 100% parity and >= 1.05x speedup.

Amdahl CandidateScanner

AST Complexity × sys.monitoring Runtime Hotspot Ranker

Scores candidate functions by log1p AST loop/math depth weighted by live sys.monitoring execution time, ignoring cold CLI scripts to focus compilation where wall-clock time is actually spent.

Zero-Fixture Tracer

Automatic Live (args, kwargs, output) Golden Trace Capture

Intercepts real production or benchmark calls via RuntimeTracer to capture serialized input/output golden fixtures automatically without requiring hand-written unit test harnesses.

Sandbox Native Builder

Isolated Native Extension Synthesis & perf_counter_ns Gate

Compiles typed native .so candidates in a temporary build directory and benchmarks them against pure Python using high-resolution perf_counter_ns timers.

FunctionSwapper + arm64

100% Output Parity Gate & macOS Inode-Safe Hot Swap

Verifies byte-exact output parity across all captured traces, unlinks before writing on macOS arm64 to prevent kernel SIGKILL code-signing traps, and rolls back automatically on any mismatch.

Before & After Engineering AuditAutonomous Python-to-Native Optimizer100% golden trace parity gate · >= 1.05x verified speedup floor

sys.monitoring hotspot profiler & parity-verified .so swap vs. static AST guessing

Before: Static AST guessing on cold code & unverified native rewrites

Static analyzers waste time optimizing cold CLI functions, and unverified native codegen can silently corrupt numerical outputs or trigger macOS arm64 code-signing crashes.

# Naive static rewrite (no runtime profile, no parity gate)
# Rewrites cold 300-line config parser; introduces C int64 overflow on edge case
After: Amdahl runtime profiler, live golden trace capture & 100% parity .so swap

Velocicode weights AST complexity by live sys.monitoring call time, harvests real golden traces, and only hot-swaps compiled .so extensions that pass 100% trace parity and >= 1.05x speedup.

# Velocicode CandidateScanner -> RuntimeTracer -> FunctionSwapper
hotspot = scan_amdahl(ast_score * sys_monitoring_ns) # targets hot inner loop
if verify_parity(compiled_so, golden_traces) == 1.0 and speedup >= 1.05:
    atomic_unlink_and_swap_arm64(fn, compiled_so)
CV & VLM rubric shift: 0.7625 → 0.953 (+0.1905 · Sharpness +55.7% · PSNR 44.88 dB)
Grounding gate: Negation parity & evidence verified

Interactive walkthrough

Step through how Velocicode works without creating an account or installing local dependencies.

Entry & problem context

Maturity: In active development

Rewriting slow Python loops in C or Rust by hand takes days of debugging, and unverified auto-generated native code can silently change program results or even run slower.

Automated Python optimization engine that pinpoints the slowest functions in a codebase, compiles them into native machine extensions, and only keeps builds whose output matches the original exactly.

Shared platform architecture sandbox

Test the live cross-language single source of truth shared across TypeScript, Python, and Rust.

Evidence grounding & negation parity check

Grounding tier: RejectedPolarityFlip (grounded: false, overlap: 1)
Socrates epistemic decision: clarify (confidence: 0.425)

Dual-gate write receipt verifier (DAWRI) & FTT math healer

Unbacked claim stripped: true
Sanitized output: No profile changes were committed in this turn.
FTT verbatim math healer: EV 142, σ 133.5515, gist positive_ev_with_downside
Hick-Hyman EIG chips: Remote or hybrid? · Target salary floor? · Company stage? (lockout: false)

Semiotic Brand Identity & Federated SoC Governance

Peircean Brand Mark, Multi-Size Squint Test & 3-Tier SoC Boundary

SoC tier: oss_public16px rubric: 96.7% (PASS)
16px · 32px · 48px squint test
Motif: latency_bolt
Iconic sign: Code terminal brackets (`< >`) bisected by a supersonic velocity lightning vector
Indexical cue: Sub-millisecond Rust/WASM execution benchmark and throughput telemetry
Symbolic archetype: Electric cyan speed and high-frequency systems engineering
SoC Boundary & Epistemic Retrieval Guardrails
Visibility & ceiling: public_oss (max tier: oss_public)
Boundary policy: Default fail-closed public OSS governance profile.
4-gram shingle novelty: 100.0% (non-redundant >15%) · Retired symbol trap: Clear (0 retired symbols)
GitHub Profile & Source Activity

Engineering Velocity & Open-Source Hygiene

Synced from @brbrainerd and @vox-foundation via the GitHub GraphQL API (GET /api/github/stats).

GitHub Profilevox-foundation/vox
Annual GitHub activity
13,842+
3,229 public · 10,136 private revisions
Curated portfolio history
19,523+
Revisions across 32 curated projects
Pull requests & reviews
463 PRs
Merged across 70 first-party codebases
Velocicode historyPython
5 revisions
31 tracked source files · Updated 2026-10-14
Language distribution across curated projectsWeighted by revision history & source volume
TypeScript38.9% (12 projects)
Rust36.9% (1 project)
Python11.3% (8 projects)
JavaScript4.7% (2 projects)
Lua4.1% (1 project)
Public open-source codebases on GitHub
vox-foundation/voxRust · 7,200 rev

Single-file full-stack programming language and Rust compiler that generates database schemas, server routes, and typed clients from one declaration.

#programming-language#rust#compiler#full-stack#lsp
brbrainerd/govSimTypeScript · 435 rev

Interactive political economy simulator with ECS civic rights, institutional modeling, policy levers, and multi-regime scenarios.

#simulation#political-economy#typescript#ecs#strategy-game
brbrainerd/neurogeckoHTML · 6 rev

Interactive 3D macaque brain atlas and stereo surgical view planner rendered in browser WebGL.

#neuroscience#webgl#threejs#brain-atlas#3d-visualization
brbrainerd/brFrameTypeScript · 23 rev

Single-binary Rust and React kiosk shell for Raspberry Pi digital photo frames with remote mobile upload.

#rust#raspberry-pi#kiosk#embedded#react
Browse all projects