Hacker News: martinrolph

New comment by martinrolph in "Embarrassingly simple self-distillation improves code generation"

martinrolph — Sun, 05 Apr 2026 08:09:41 +0000

Think of it less like a test suite and more like an exam. If you're trying to differentiate between the performance of different people/systems/models, you need to calibrate the difficulty accordingly.

When designing a benchmark, a pass rate of roughly 50% is useful because it gives you the most information about the relative performance of different models. If the pass rate is 90%+ too often, that means the test is too easy: you're wasting questions asking the model to do things we already know it can do, and getting no extra information. And if it's too low then you're wasting questions at the other end, trying to make it do impossible tasks.

New comment by martinrolph in "Endgame – A short story about machines optimising for engagement"

martinrolph — Thu, 30 Oct 2025 13:53:14 +0000

I wrote a stream of consciousness that turned into a piece of flash fiction. A peek just a little way into the future, inspired by the vertigo I felt on the announcements of Vibes and Sora 2, the AI-only short form video feeds. What happens, I wonder, when the AIs that are optimising your For You page for addiction achieve their goals? What is the endgame for addictive content?

Endgame – A short story about machines optimising for engagement

martinrolph — Thu, 30 Oct 2025 13:53:14 +0000

Article URL: https://malagostudio.substack.com/p/endgame

Comments URL: https://news.ycombinator.com/item?id=45760042

Points: 1

# Comments: 1

New comment by martinrolph in "Overlock – IoT Exception Tracking"

martinrolph — Thu, 16 Nov 2017 15:43:14 +0000

In an IoT deployment, an issue that is detected on one device (e.g. an exception on the server) may actually originate on an entirely different device (e.g. the end device sends some incorrect data). Overlock collects together the state of all the devices involved in a given request and the associated logs in one place so that you can track the error back to where it originated.

New comment by martinrolph in "Offload: Build seamless multiscreen interfaces"

martinrolph — Tue, 04 Nov 2014 14:34:33 +0000

When researching this I came across a few tools for using a phone as a trackpad, for example:

https://play.google.com/store/apps/details?id=com.thingsstuf...

We couldn't find anything which worked as a digitiser and that is certinaly something we're considering.

New comment by martinrolph in "Offload: Build seamless multiscreen interfaces"

martinrolph — Tue, 04 Nov 2014 13:41:29 +0000

Oops! Sorry, changed the title.

Offload: Build seamless multiscreen interfaces

martinrolph — Tue, 04 Nov 2014 12:13:18 +0000

Article URL: http://offload.strikingly.com/

Comments URL: https://news.ycombinator.com/item?id=8556140

Points: 44

# Comments: 9

Ask HN: Are there any web-apps aimed at e-readers?

martinrolph — Mon, 03 Nov 2014 20:53:41 +0000

It would be great (for me at least) to be able to do several common activities on my Kindle. The main motivator is that I find it hard to sleep if I've been looking at screens late at night. For example, I'd love to be able to use some kind of messaging client.

I was never able to get Kindle Active Content (the closest thing there was to an app store) as I'm not US-based and that seems to be disappearing anyway.

Which means the only available option is using web-apps. Of course, the obvious problem is that they're totally unoptimised for e-ink screens where scrolling, colour and movement are all either poorly implemented or not available at all.

Are there any web-apps designed to work well with e-readers?

Comments URL: https://news.ycombinator.com/item?id=8553417

Points: 2

# Comments: 1