feat: add Linux Kernel Builder benchmark mode

[HeyGarrison]

Summary

• Add a new linux-kernel-builder benchmark mode for LKB (Linux Kernel Builder) to measure sustained sandbox performance on a real workload (defconfig + bzImage).
• Introduce a dedicated provider cohort, LKB scoring, terminal table output, and JSON result writer under src/linux-kernel-builder/.
• Wire LKB into src/run.ts, add npm scripts for full-run and per-provider runs, and include workload metadata (linux-kernel-builder, LKB) in results.
• Add a dedicated GitHub Actions workflow at .github/workflows/linux-kernel-builder-benchmarks.yml that runs on push/PR for relevant path changes.
• Fix command execution to run the benchmark as a multiline shell script (resolving the prior bash then; syntax failure).

Validation

• Ran npx tsc --noEmit
• Ran npm run -s bench -- --mode linux-kernel-builder --provider namespace --iterations 1 (exec path verified; provider skipped without local credentials)

[github-actions]

Storage Benchmark Results

10MB Files

#	Provider	Score	Download	Throughput	Upload	Status
1	Tigris	95.9	0.10s	810.7 Mbps	0.58s	10/10
2	AWS S3	93.6	0.12s	697.8 Mbps	0.36s	10/10
3	Cloudflare R2	82.9	0.39s	215.7 Mbps	0.92s	10/10

---

View full run · SVGs available as build artifacts

[github-actions]

Sandbox Benchmark Results

Sequential

#	Provider	Score	Median TTI	P95	P99	Status
1	daytona	98.2	0.11s	0.29s	0.29s	10/10
2	e2b	96.2	0.35s	0.44s	0.44s	10/10
3	vercel	95.9	0.38s	0.46s	0.46s	10/10
4	blaxel	95.4	0.44s	0.50s	0.50s	10/10
5	hopx	87.7	1.10s	1.41s	1.41s	10/10
6	modal	84.1	1.40s	1.87s	1.87s	10/10
7	runloop	84.0	1.38s	1.93s	1.93s	10/10
8	cloudflare	81.7	1.53s	2.27s	2.27s	10/10
9	namespace	74.2	1.82s	3.74s	3.74s	10/10
10	codesandbox	68.9	2.62s	3.84s	3.84s	10/10

Staggered

#	Provider	Score	Median TTI	P95	P99	Status
1	daytona	98.0	0.13s	0.30s	0.30s	10/10
2	blaxel	95.5	0.43s	0.48s	0.48s	10/10
3	vercel	95.2	0.37s	0.65s	0.65s	10/10
4	e2b	94.3	0.33s	0.92s	0.92s	10/10
5	hopx	88.7	1.09s	1.19s	1.19s	10/10
6	namespace	81.3	1.73s	2.06s	2.06s	10/10
7	modal	81.3	1.56s	2.33s	2.33s	10/10
8	runloop	80.6	1.88s	2.04s	2.04s	10/10
9	cloudflare	73.9	1.75s	3.91s	3.91s	10/10
10	codesandbox	62.3	3.63s	3.98s	3.98s	10/10

Burst

#	Provider	Score	Median TTI	P95	P99	Status
1	daytona	98.0	0.13s	0.31s	0.31s	10/10
2	blaxel	94.9	0.47s	0.56s	0.56s	10/10
3	e2b	94.6	0.46s	0.65s	0.65s	10/10
4	vercel	87.8	0.35s	2.52s	2.52s	10/10
5	hopx	82.1	1.71s	1.92s	1.92s	10/10
6	modal	81.5	1.71s	2.06s	2.06s	10/10
7	namespace	79.0	1.92s	2.36s	2.36s	10/10
8	runloop	68.3	2.97s	3.47s	3.47s	10/10
9	cloudflare	63.4	2.22s	5.81s	5.81s	10/10
10	codesandbox	60.1	3.45s	4.79s	4.79s	10/10

---

View full run · SVGs available as build artifacts

[github-actions]

Linux Kernel Builder Benchmark Results

#	Provider	Score	Build Median	Build P95	Build P99	Status	Sample Failure
1	blaxel	0.0	--	--	--	0/1	Command failed with exit code 2: Missing downloader: curl or wget
2	daytona	0.0	--	--	--	0/1	Command failed with exit code 2: linux-6.12.24.tar.xz: OK tar: Unexpected EOF in archive tar: Unexpected EOF in archive
3	e2b	0.0	--	--	--	0/1	Command failed with exit code 3: linux-6.12.24.tar.xz: OK Missing dependency: bison
4	hopx	0.0	--	--	--	0/1	Command failed with exit code 124: Command timed out after 30 seconds linux-6.12.24.tar.xz: OK
5	namespace	0.0	--	--	--	0/1	Command failed with exit code 1: time="2026-04-16T02:09:39Z" level=fatal msg="exec failed with exit code 3" linux-6.12.
6	runloop	0.0	--	--	--	0/1	Command failed with exit code 2: tar (child): xz: Cannot exec: No such file or directory tar (child): Error is not recov

---

View full run · JSON artifacts available in the run artifacts