hanzo.ai/app/(marketing)/benchmarks/sandbox/page.tsx

/**
 * /benchmarks/sandbox — time to a running sandbox, by primitive.
 *
 * The table renders the sandbox pass of lib/data/benchmarks-runtime.json: each
 * primitive with a Hanzo figure that no figure published beside it beats.
 */
import { phrase } from '@hanzo/i18n'
import { T } from '@hanzo/i18n/react'
import { pageMeta } from '@/lib/page-meta'
import { BENCHMARKS, bench } from '@/lib/benchmarks'
import { BenchJsonLd } from '@/components/benchmarks/JsonLd'
import { Grid, H2, Meta, More, Note, P, Shell, Spec, type Col } from '@/components/benchmarks/kit'
import { RUNTIME, shots, type Shot } from '@/lib/bench-data'

const B = bench('sandbox')!

export const metadata = pageMeta({
  title: 'Sandbox cold start benchmark',
  description:
    'Time to a running sandbox: a V8 context ready to execute JavaScript in 0.13 ms, the median of one pass, reproducible by one command.',
  path: '/benchmarks/sandbox',
})

const MEASURES = ['context start latency']

/* What each primitive can RUN is particular to this page, so it is joined on
   here rather than carried in the shared rows. */
const RUNS: Record<string, string> = {
  'V8 context': phrase('JavaScript, sharing the heap'),
}
/* The primitives shown: each a Hanzo measurement that no figure published beside it beats (lib/bench-data). */
const ROWS = shots('sandbox')
const cols: Col<Shot>[] = [
  { head: <T>primitive</T>, get: (r) => r.what, left: true },
  { head: <T>measured here</T>, get: (r) => r.ours },
  { head: <T>what it can run</T>, get: (r) => (RUNS[r.what] ? <T>{RUNS[r.what]}</T> : '–'), left: true },
]

export default function Page() {
  return (
    <>
      <BenchJsonLd b={B} measures={MEASURES} />
      <Shell b={B}>
        <Spec b={B} />

        <H2><T>What is measured</T></H2>
        <P>
          <T>{'The time from asking for a sandbox to having one that will execute code. A V8 context is a fresh global inside a V8 isolate that is already running: it is cheap because it shares that isolate’s heap. Timings are the median of the pass.'}</T>
        </P>

        <H2><T>Results</T></H2>
        <Meta>
          <T values={{ machine: RUNTIME.passes.sandbox.machine }}>{'{machine} · the pass of 2026-09-07 · median of the pass'}</T>
        </Meta>
        <Grid cols={cols} rows={ROWS} keyOf={(r) => r.what} />
        <Note>
          <T>{'A context runs JavaScript, which makes it the right primitive for a JavaScript transform. An agent that installs a package or runs a shell gets a microVM instead, which is a separate measurement.'}</T>
        </Note>

        <H2><T>What this measures, and what it does not</T></H2>
        <P>
          <strong>
            <T>Cold start is not throughput.</T>
          </strong>{' '}
          <T>
            {'This is the time to the first instruction, on an idle machine. It says nothing about what happens when a thousand sandboxes are asked for at once.'}
          </T>
        </P>
        <P>
          <strong>
            <T>One machine, one day.</T>
          </strong>{' '}
          <T>
            {'A single pass on one laptop, reproducible by one command. Read it as a floor that has been established rather than as a service level.'}
          </T>
        </P>
        <P>
          <strong>
            <T>It is one of three costs, not the cost.</T>
          </strong>{' '}
          <T
            values={{
              dormant: <a href="/benchmarks/fleet"><T>477 bytes and a 0.034 ms resume</T></a>,
              live: <a href="/benchmarks/goroutine"><T>601 bytes of heap and 125 ns to spawn</T></a>,
            }}
          >
            {'A sandbox is what an agent needs to run a command. What it costs to keep an agent that is doing nothing is {dormant}, and what a running agent holds is {live}. The three are separate measurements with separate commands and they do not add up into a single number.'}
          </T>
        </P>

        <More b={B} all={BENCHMARKS} />
        <Meta>
          <T
            values={{
              harness: <a href={B.harness.url}>{B.harness.repo}</a>,
              script: <code>sandbox.mjs</code>,
              start: <code>colima start</code>,
            }}
          >
            {'harness and the committed table: {harness} · {script} needs a container runtime; {start} is enough on macOS'}
          </T>
        </Meta>
      </Shell>
    </>
  )
}