Straight Answers
An LLM assessed 896 question → answer exchanges, the host's own answers included. Speaker names were hidden during assessment. Each exchange was marked answered, partly, redirected or not addressed, with a quote as its source. Every exchange below is timestamped and plays. How it works →
These readings are observational: the transcript shows whether the question was answered directly, whatever the reason. It is an AI reading of a public exchange, never a claim about intent, and the tape is one tap away on every row. full disclaimer →
Every one of the 168 people on the record is accounted for: 2 carry a full rate (8 or more assessed questions on raw tape), 161 more have their questions counted (33 of them with at least one not answered directly; too few raw questions for a fair percentage, edited feed included), and the remaining 5 never took a direct question in any tape we hold: compilations, panels and cameo appearances leave nothing to assess.
No needle for Erik Bernhardsson: fewer than 8 assessed questions on raw tape, and a percentage on a sample that small would mislead. Their counts are real: 2 not answered directly in 6 assessed questions.
Erik Bernhardsson: every exchange, playable
The question asked, then the AI assessment note from the answer. Tap to play the moment. Not addressed = the question was not engaged; redirected = acknowledged, then steered elsewhere.
“So a lot of folks I know, um, may use one of the existing vector DBs, or in some cases are just using Postgres with, um, with PG vector, right? How do you think about the need for vector databases as sort of standalone Pieces of infrastructure versus just, you know, adopting Postgres versus doing something else.”
answered: “that's actually a bigger question that I'm more excited about”
“You know, many of them actually, uh, work across multiple. How do you think about that in the context of modal in terms of your own compute versus hyperscalers versus, you know, the ability to run anywhere?”
answered: “And of course, there's also a sort of security compliance aspect of this.”
All speakers
All 2 people with 8 or more assessed questions on raw tape. Click a name to see their exchanges. highest rate none identified small samples
| Person | Not answered directly (%) | Redirected + not addressed | Questions | Argument clarity /5 |
|---|---|---|---|---|
| Elad Gil Founder, Gil Capital | 30% | 3 | 10 | 4.0 |
| Mustafa Suleyman CEO, Microsoft AI | 0% | 0 | 9 | 4.5 |
Percentages come from raw unedited recordings only: produced podcast audio has tangents and stumbles cut in the edit, which moves the speaker up by about 12 percentile points. The small-sample counts include the produced feed. This is a reading of a transcript. It is not an accusation of dishonesty, and there are many good reasons not to answer a question directly (confidential numbers, unreleased products, someone else's news to break).