Systima's Growth Play: Name the Market Leader in Your Headline and Back It with a Reproducible Benchmark, Not an Opinion.

The Play

Systima didn't publish "we compared two coding agent harnesses."

It published "Claude Code Is Way More Token-Hungry Than OpenCode. We Measured Exactly How Much" — naming the market leader directly in the headline, then backing every claim with a number: character counts, token counts, cache writes, pass rates.

That combination — a named incumbent plus a reproducible measurement — is what turned an internal benchmarking exercise into a front-page Hacker News thread and multiple syndicated write-ups.

Why it worked

A post titled "AI coding agent token comparison" is invisible. A post that names the tool your reader already uses, makes a specific claim about it (roughly 4.7x more tokens on a first request, over 50x more cache writes on one task), and shows exactly how that number was produced is impossible to scroll past — especially for the exact audience, developers running these harnesses daily, who can verify or dispute it themselves.

The methodology mattered as much as the headline. Systima reported concrete inputs — a 27,344-character system prompt, 99,778 characters of tool schemas, a "10-lane test suite" run "5 runs each" — instead of just asserting a winner. That specificity is what let the post survive the inevitable pile-on of readers trying to poke holes in it.

Systima also didn't cherry-pick the result. On one multi-step comparison, Claude Code's habit of batching tool calls into fewer requests brought its cumulative total below OpenCode's, even though Claude Code's baseline overhead was higher. Including that nuance — a place where the "loser" of the headline ratio actually came out ahead — is what kept the piece reading as research instead of a hit piece.

The growth play to steal

1. Pick a comparison your exact audience already argues about informally, then be the first to publish real numbers instead of another opinion piece.

2. Name the incumbent directly in the headline. Vague framing ("Comparing two popular tools") gets none of the search or share traffic that naming gets.

3. Publish your methodology, not just your conclusion — exact inputs, exact test conditions, exact run counts — so technical readers can verify or reproduce it instead of dismissing it.

4. Report the case where results are close or where your own approach loses ground. A benchmark that never shows nuance reads as marketing, not research.

5. Expect and prepare for the comment-section fight. A benchmark that names a market leader will get scrutinized in the replies; that scrutiny is what proves the post was rigorous enough to be worth arguing about, and it's what keeps the thread — and your traffic — alive.

Bottom line

Systima's post spread because it did the unglamorous work most comparison content skips: it named names and showed the exact measurement behind the claim. That's the transferable move — pick the argument your audience is already having, publish the receipts, and let the incumbent's name do the SEO and social work a generic title never could.

Source: https://systima.ai/blog/claude-code-vs-opencode-token-overhead