NAD C315BEE amplifier buzzing

I’m sticking this up in the off chance it’ll turn up in someone’s Google search one day and save them some time or money.

If you found this you probably did what I did which was forget to turn off your NAD 315BEE amplifier. And then one night, at 3am, you woke up to find your speakers buzzing REALLY LOUDLY. Even when you put the amp into standby, which is odd.

The two large 10000μF smoothing capacitors have blown and need replacing. If you’ve got a soldering iron and stable hand you should be able to do it yourself.

If you’re in the UK RS stock these replacements which are the right specification and size. You can find guides on YouTube for how to swap them.

Internet Research #1

1. I’ve been enjoying noodling around with some electronics recently. One of the things I’m working on is converting a pair of beautiful old Bang and Olufsen CX50 speakers into an open source Sonos-esque multiroom audio system. I’m following HifiBerry’s guide for the hardware part, and in it there’s two 3D printed parts you need.

So I go on 3D Hubs, upload the two .stl files provided, get a quote, credit card, and now it’s printing somewhere in Farringdon. You still need to know whether you want PLA or ABS or whatever, but I was very impressed at how easy the whole thing was. Just another mundane e-commerce transaction. We’re out the other side of the hype cycle.

2. Simon Willison linked to libpostal on Twitter. On the surface, a useful library for turning unformatted addresses into structured data. Useful, but nothing out of the ordinary. But on a closer look:

libpostal’s international address parser uses machine learning (Conditional Random Fields) and is trained on over 1 billion addresses in every inhabited country on Earth.

Increasingly, just chunking loads and loads of data at a problem becomes an option. Why bother writing a parser to handle every condition and edge case if you can just throw all the known address data in the world at it and let the computer work it out? The writeups on how it works are interesting and thorough.

Statistical NLP on OpenStreetMap: Part 1, Part 2

libpostal screenshot

3. Mapzen funded libpostal, and loads of other brilliant geodata projects. Sadly they’re shutting down, but because it’s all open source the tens of £millions of value won’t be lost. Their migration guide walks through the alternatives to all their hosted services.

Aaron’s post about Who’s On First, the gazetteer of places, is a brilliant example of how to ensure an important project lasts beyond the lifespan of a single organisation or corporate strategy. Lots of loosely joined pieces, many easy to self or community host. Designing your project to be defended against the organisation that hosts it means designing that in from day one.

In many ways everything about the way Who’s On First has been designed has been done with this day in mind. We all endeavour to achieve the sort of “escape velocity” that immunizes us from circumstance but that is rare indeed and there was always a chance this day would come. So while “success” was the goal in many ways preventing what I call “the reset to zero” has always been of equal importance.

Who’s On First, Chapter 2

DRAWING, THE CHROMO CUBE, 1980

4. I think spotted this on FaveJet probably via Russell.

David Rudnick’s beautiful drawings of MiniDiscs. Drawings. In Photoshop.

Just give Jason some money

This is the most important thing. Give Jason some money. You want to live in a world where people like Jason can make a living writing on the internet. The way to get that is to give him some money.

Do it.

Generating English village names with neural networks

Roll 13/02

I’m trying to wrap my head around the new generation machine learning tools: deep neural networks and the like. It feels like this technology is approaching where databases were 20-30 years ago: the tooling is getting easy enough that an idiot like me can have a stab at wiring something up, even if I don’t quite understand all the magic incantations that I need to type. And it’s pretty clear it’s going to be important.

The world seems to be settling on Tensorflow, for now, so I had a go at getting something stupid up and running. I ended up making an English village name generator, using a corpus from OS Open Names (with a healthy amount of awk and grep), and a character level recurrent neural network written in Tensorflow.

Like I was with SQL many moons ago, I think I understand some of the principles, what’s possible and what’s not, and I can make sense of someone else’s code – but it’s a bit of struggle getting all the words in the right order when I have to change anything.

Anyway, the results are quite fun – here’s 20 of them:

Allers Bottom
Culack
Swrarby
Fenwall St Eastake
Anbarth
St Ninhope
Thawkanham Water Green
Mige Lane
Up Maling
Firley Dinch
Lindlemere
Stan Hill
Hiddlesley
Pibley
Hunmastreet
Shenworth Strough
Hendrelds Hill
Scottedane
Crickines
Stranal Footh

And here’s another 980.

Update: I turned this into a Twitter bot, @urnowentering.

FaveJet

I’ve used Jason Kottke‘s Stellar for a couple of years now, and love it. If you’ve never tried it, it’s a stream of things your friends are favouriting, mostly on Twitter, but YouTube, Vimeo and other stuff too.

Favourites are quieter than retweets, and more nuanced. A little nod of acknowledgement, a gentle arm around a shoulder, or sometimes just a smile.

Stellar surfaced those in a way that always felt like a superpower: the ability to see a little further, beyond my immediate network, slicing through the fog.

And so when it went offline a couple of months ago, Twitter got a little less enjoyable for me.

To plug the gap, I’ve made a little thing called FaveJet, and you can have a go if you like.

FaveJet

Running this type of thing is hard. Pulling down everyone’s favourites every 30 mins and reconciling them into an activity stream is intensive as websites go. It seems fine for the 40 people using it at the moment, but it’ll look quite different at a thousand or five thousand.

But I’d like anyone to use it, and I don’t want it to be beholden to the whims of me and my rare free time. I’d like to find a model to support it, that doesn’t involve emptying my pockets alone, or losing precious family time.

Maybe that looks like a company with some people paying and using those proceeds to pay for work on it, or maybe that looks like some kind of co-ownership/co-op model, splitting the work between a few people. Or maybe a mix.

I’m not sure, but I’d like to find something that works, and hopefully an approach I can use to support other niche products in the future.

If you’ve got any ideas, you’d like to help or you’ve tried this in the past, please drop me a line.

Anyway, in the meantime, feel free to give it a go. It only works for Twitter at the moment, and I’ve capped sign ups to 20 per day for now. Happy faveing.

Offset, Onwards

Offset

I’m leaving Offset at Christmas.

It’s been fun, hard work, with a wonderful team, backed by a great organisation. I joined to invent and launch some stuff, and the course is charted to do just that. The rest of the year will be busy as we ready for a launch, and then it’s time to hand the baton on.

I’m not sure what’s next yet. I’ve got a few days spare before Christmas, and from January I’m looking for the next thing(s). So, if you think you need some product invention/prototyping/engineering/reckons, drop me a line.

VEL-O-DROME

I went to the Six Day at the Olympic Velodrome this week. Great atmosphere, lots of fun.

They’ve gone full Wipeout with it all. A DJ from Ministry of Sound watching the action and keeping tempo accordingly. Coloured flood lighting. Introducing all the riders one by one as they do laps. Very little hanging around between events, and you mostly knew what was going on.

And whoever had the Derny riders come out to Ride of the Valkyries deserves a bonus.

Velodrome

Derny Racing

Camera

Hey Thermostat

Alright then, that works.

Siri

I promise I’m never blogging about my thermostat ever again.