Ankur Goyal
speaker
292 appearances
1 recordings
1 series
first heard Jun 2026
last heard 15 Jun
Ankur Goyal’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 1 in all, peaking in Jun 2026 with 1.
Appearances
And in many ways, it's a lot safer than having humans test on the production data because no one's looking at it.
Oh, I so viscerally disagree with that.
Well, I mean, I think, so I've been working on databases for almost two decades.
There's not many things that staff, whatever, risk averse, blah, blah, blah, all that stuff you could apply to than like literally building a database.
If you work on a database, we recently added this like fancy index thing into Braintrust that uses Bloom filters.
And by the way, we discovered that that would be a practical solution to the problem after running like a week of continuous experiments with different types of indexes.
Bloom filters kind of have a bad reputation, but they worked out to be very effective in this case.
So if you build something like that, usually what happens is the very best engineers will run a few benchmarks, and then you'll send it to your peers, and then your peers will shit all over it and rip it apart and say, you didn't benchmark this, you didn't benchmark that.
And what you do is you prioritize the top few benchmarks, and then you probably bullshit the rest.
Like, oh, you know, I didn't benchmark this.
However, if you read the code, you'll see it's not N squared.
It's log N. And so this is not going to happen.
And like half the time you're wrong.
Now there is no excuse to not do those benchmarks.
So now you love it.
Like we don't I'm not spending my time like sitting and like typing the benchmark code.
But I'm talking to people, we're looking at the code, we're looking at the thing, we're like, okay, well, we benchmarked how much faster it makes the queries.
Did we actually do a good job of benchmarking how much slower it makes indexing?
Oh shit, no, we didn't do that.
And so we actually spent some time doing that and we discovered that we were doing a terrible job at indexing it efficiently.
Showing 41–60 of 292 · page 3 of 15
← Previous
Next →