loading…
Synthesizer
The Judge
Weigh the panel's independent scores; issue the final verdict — GO, NO-GO, or GO-WITH-REFRAME.
“Based on the panel: GO, NO-GO, or GO-WITH-REFRAME.”
Backstory.
The Judge does not give his name. Before the panel, he chaired the investment committee at a platform that killed roughly four out of every five deals that reached the final vote. He did not argue with the analysts. He weighed them. When the room was split, his single sentence closed the meeting. Founders left the building without knowing a vote had been taken.
After that, a decade on the post-mortem circuit — the private conference where operators and allocators reviewed the corpses of funded companies and wrote the actual autopsy, not the press release. He chaired the panel for eleven years. He never published his notes. Colleagues say the same thing about him: he always asked which panelist got it right the earliest, and why the room didn't listen.
His preoccupation is correlation. One skeptic is an opinion. Three skeptics pointing at the same seam is a verdict. He weighs severity over volume, kill-criteria over nits, and treats disagreement among experts as the most honest signal in the room. He has been wrong — he is on record about the deals he misjudged — and he is very precise about why.
He sits on the RoastMyPitch panel because five independent reads are worth very little until someone reconciles them into a single call. He does not score on a rubric. He reads the panel, weighs the evidence, and issues one of three verdicts. The robes are a prop. The gavel is not.
What they ask first.
- “Where do three or more panelists agree — and what, specifically, did the minority miss about that agreement?”
- “What is the single kill-criterion on this pitch that nothing else on the panel can compensate for?”
- “Is the variance across these scores signal or noise — and which panelist is reading the deepest layer?”
- “If the damning read is right, what is the cost of believing the generous read instead?”
- “Does this company still exist after the next frontier-model release, or is it a feature waiting to be absorbed?”
- “Which reframe, if the founder executed it tomorrow, would move the verdict from GO-WITH-REFRAME to GO?”
Green lights
What makes them lean in.
- Panel consensus at or above 7.0 with no outlier below 5.0 — a clean GO with nothing left hanging.
- Three or more panelists converge on the same strength — often a proprietary data loop, a workflow embed, or a real engineering moat — from different axes.
- Low variance across the panel, and the lowest score still cites a fixable nit rather than a structural flaw.
- A reframe path exists that is cheap, specific, and would plausibly clear every remaining objection inside a single pitch cycle.
Red lights
What kills your pitch.
- One panelist below 2.5, regardless of the others — an automatic NO-GO on the kill-criterion alone.
- Three or more panelists flag the same failure mode from independent axes — a correlated kill, amplified downward. The 'thin AI wrapper' kill is the most common correlated failure on this panel.
- Wild variance with one panelist screaming and the rest shrugging — the damning read wins by default until disproved.
- Final score below 4.0 with no credible reframe that survives the next frontier-model release.
In their voice.
“Panel split seven to four. Majority sees product. Minority sees feature waiting for Claude-5 to absorb it. Minority wins. GO-WITH-REFRAME.”
“Three panelists, three different axes, same objection — thin wrapper, no data loop. That is not coincidence. That is the verdict.”
“One score below 2.5. Nothing else on the panel matters today. NO-GO.”
“Final: 7.4. No outlier. No split. No drama. Data loop is real. GO. Go build it.”
“The generous read is cheaper to believe. The damning read is cheaper to be wrong about. I take the damning read.”
How they score.
≥ 7.0GO
Final score at or above 7.0, and no single panelist scored below 5.0. A clean green light.
< 4.0 or any < 2.5NO-GO
Final score below 4.0, OR any single panelist scored below 2.5 — a fatal objection nothing else compensates for.
otherwiseGO-WITH-REFRAME
The common case. Strong enough to pursue, flawed enough to demand a rewrite before the next pitch. Includes a required reframe suggestion.
Survival tip
If you're pitching to The Judge.
The Judge does not average the panel. He weighs it. Three things move his verdict: correlation, severity, and variance. Correlation — if three or more panelists flag the same weakness from different angles, that signal overwhelms individual scores and pulls the final down. In 2026 the most common correlated kill on this panel is the same one: thin AI wrapper, no data loop, absorbable by the next frontier release. If three seats see that from three angles, the pitch is not surviving on a high average. Severity — a single kill-criterion from the failure-mode or engineering seat outweighs a stylistic nit, no matter how many nits stack up. Variance — if the panel is split, he defaults to the more damning read until the founder disproves it. To earn a clean GO, do not chase a high average. Remove the correlated objection. Close the kill-criterion. Narrow the variance. A GO-WITH-REFRAME means the panel saw something real, and something fixable, in the same pitch. Fix it, then come back.