01 · Roasts
Commit History? What Commit History?
29 public commits in a year, and half of them appear to be 'initial commit' fired within 44 seconds of repo creation. jvmes is not iterating — jvmes is teleporting finished codebases into existence.
The Blitz Architect
peanut (OS), celene (CPU + toolchain), myotis (SSA compiler), lunara-ml (ML compiler), gpu-kernel-suite (CUDA kernels) — all shipped in single-session bursts. Impressive architecture, zero commit history. git log reads like a greatest-hits album with no B-sides.
Abandoned Lunara Prime
You built lunara as an ONNX→PTX compiler scaffold, apparently decided it was too clean, then rebuilt it as lunara-ml. The original lunara sits there with no README and a single commit, a haunted house for your ambitions.
cryogen: Dreams Written in Markdown
'Vulkan ray tracing, physics, networking, ECS' — all described in the README, none of it visible in source files, entire repo created in 37 seconds. This is a feature roadmap cosplaying as a codebase.
0 Stars Across 13 Repos
totalStars = 0. Every single one. You're building operating systems and GPU compilers in a sealed vault. At some point the portfolio needs a portfolio.
Built using
Zoral
Shadows one worker for a week, then takes over their job with zero extra setup. Behaves exactly like the original.
zoral.ai
02 · Category breakdown
- Impact25% weight56D
- Consistency20% weight55D
- Quality20% weight57D
- Depth15% weight50D
- Breadth10% weight65C
- Community10% weight50D
03 · Stats
365-day commit heatmap
25 active days
Language distribution
- C++48%
- Makefile23%
- Python12%
- C5%
- TypeScript4%
- CMake3%
- Other5%
04 · Numbers
Owned repos
non-fork
11
Commits
last 12 months
29
Followers
22
Joined GitHub
Nov 2022
05 · Top repos
jvmespark /
lunara-ml
Early-stage ML compiler IR and CUDA codegen framework with typed C++, documented architecture, comprehensive test suite, and working CPU/GPU backends—but no README, zero adoption, and nascent feature set.
jvmespark /
peanut
A bare-metal 32-bit x86 hobby OS with bootloader, preemptive multitasking, SMP, paging, buddy allocator, and embedded CPU emulator. Typed C codebase with structured architecture and comprehensive design docs, but experimental scope and zero public adoption.
jvmespark /
myotis
Early-stage SSA compiler in C++20 with front/mid/back pipeline; lexer, parser, IR builder, dominators, and CI working; multiple stub optimizations not yet implemented.
jvmespark /
celene
Complete 8-bit CPU stack (ISA, assembler, compiler, emulator, Verilog) with typed Python, structured layout, tests, and working examples. Fresh project (1 day old, 1 commit) with zero adoption but ambitious scope demonstrates sustained design work.
jvmespark /
gpu-kernel-suite
Educational GPU kernel reference implementations (CUDA/Triton/Pallas) for SGEMM, softmax, RMSNorm, FlashAttention, with typed Python harness, tests, and CI. Created Apr 26, 2026 with 1 recent commit; experimental single-author project.
jvmespark /
jvmespark.github.io
Personal Next.js 14 blog with MDX, D3 graph, and Tailwind. Typed, documented, well-structured for a portfolio site but brand-new (1 day old, 1 commit) with 0 stars/forks—experimental personal project.
jvmespark /
myotis2
A complete SSA compiler (lexer → parser → IR → optimization passes → code generation) for a C-like language targeting x86, ARM, and PTX. One-shot dump created hours ago; 0 stars, 78 KB, MIT-licensed.
jvmespark /
cryogen
Game engine scaffold with ambitious feature claims (Vulkan ray tracing, physics, networking) but only 91 KB, 1 commit, zero adoption. README describes architecture but no actual source files delivered yet.
jvmespark /
lunara
Single-commit ML compiler stub (74 KB, 1 recent push) with architectural documentation but no verifiable implementation, tests, or CI—appears to be a project scaffold or placeholder.
jvmespark /
jvmespark
Empty scaffold repo created May 5, 2026, containing only a README with personal bio/CV content. No source code, tests, CI, license, or meaningful project deliverables.
06 · Timeline
- Nov 27, 2022Joined GitHub
- Feb 21, 2026Created myotis — SSA compiler: parsing → IR → optimization passes → AVX-512 + PTX codegen. LLVM-style architecture, built from scratch.
- Mar 1, 2026Created lunara-ml
- Apr 26, 2026Created lunara — ML kernel compiler: StableHLO → MLIR linalg → Triton → PTX.
- Apr 26, 2026Created gpu-kernel-suite — ML Kernel implementations. CUDA, Triton, JAX, MLIR. Nsight traces, roofline analysis, annotated Triton IR.
- Apr 28, 2026Created myotis2 — SSA compiler: parsing → IR → optimization passes → AVX-512, AAPCS64, PTX codegen. LLVM-style architecture.
- May 1, 2026Created celene — 8-bit CPU + toolchain: ISA, assembler, emulator, compiler, runtime, Verilog CPU.
- May 1, 2026Created jvmespark.github.io — A personal website built with Next.js 14, MDX, Tailwind, and D3.
- May 5, 2026Created cryogen — Game engine with multiplayer networking, 3D mesh loading, rigid-body physics system, and real-time physically-based ray tracing.
- May 5, 2026Created peanut — 32-bit x86 multi-core operating system booted on bare metal via a hand-rolled two-stage BIOS bootloader.
- May 5, 2026Created jvmespark
- May 5, 2026Most recent push to jvmespark
07 · Compare
08 · Rubric
How this score was produced
Overall = Σ (category × weight) + gentle top-end curve
Tier thresholds
▸ How the pipeline works
- 01Scrape.Pull every non-fork repo pushed in the last 90 days, plus your contribution calendar, followers, and language byte counts — straight from GitHub's REST & GraphQL APIs.
- 02Triage.A small model reads every repo's file tree + README and picks the 20 files per repo that actually reveal how you code.
- 03Grade each repo. All repos run in parallel through a fast scoring model that reads the picked files and rates each one independently on Impact, Quality, and Depth — with evidence citations.
- 04Aggregate. A larger reasoning model combines the per-repo scores with server-computed stats (heatmap, commit cadence, language entropy, follower count) to produce the 6-dimension profile score + roasts.
- 05Correct.Deterministic server-side checks enforce anchor-scale floors (e.g. a profile with 2,000+ public commits can't score 30 Consistency) and recompute the final verdict.
~90 seconds per profile, ~$0.25 in compute. Total of ~240 files read across your top-12 repos. One rating per GitHub account per day.
▸ Data sources & caveats
- Heatmap & commit totals: GitHub GraphQL
contributionsCollection— covers the last 365 days, includes private repos when the user has opted in (default). - Language %: byte totals across the top 30 owned non-fork repos.
- Curve: a small upward nudge centered on raw score ≈ 70, capping at 100. Prevents specialists from being unfairly penalised for narrow breadth.
- Anchor corrections: when server-measured signals (e.g. privateWorkLikely, multiRepoVolume, follower count) mandate a minimum category score, the aggregation step enforces it. These are signal-conditional, not identity-based floors.