Dell Infuriates Me

Sunday rant: I have a 2016 Dell XPS (high-end) laptop. I bought a Dell 25″ 4K monitor. And on Friday received a Dell Thunderbolt dock for the monitor. I plug it all together and although the monitor displays wonderfully, the dock is not passing USB through. So I start fiddling around with “unplug from dock, plug into laptop, confirm the peripheral is working,” stuff. And then the laptop BSODs. Machine boots, connects to dock, everything’s fine for 30 seconds, BSOD. Start to Google. “Update your laptop BIOS.” (For a fucking docking station!). It takes a goddamn hour to find the BIOS update on the Dell Website for their goddamn flagship laptop, but whatever.

Still BSODs. Now it’s telling me that I have to update the firmware on the dock. But I cannot update the firmware because if I attach the dock to the laptop to update it, it BSODs. So, there’s this few-second window before the BSOD where I see that I have to update my Thunderbolt Driver on the laptop.

So I download the driver and run the installer for the Thunderbolt Driver. The installer doesn’t give any option other than “Uninstall.” So I say “OK, I’ll uninstall and reinstall.” I uninstall. Fine. I go to reinstall. I’m told I don’t have sufficient permission. So I run as administrator. I still don’t have sufficient permission. So I end up editing the registery to turn off user protection. (Remember, this is all for a docking station).

I now can run the “install” option, but it refuses to continue because it sees some pre-existing value in the registry. (Which I take to mean it’s “Uninstall” function didn’t actually, you know, uninstall.) It then rolls back the Thunderbolt install and leaves me with my current situation:

A half-upgraded machine with user access protection turned off, less functionality than it had before, and it still BSODs whenever I turn on the dock. All with a respected company’s flagship hardware.


Programmed my first Alexa skill: I was shocked by what I found!

Although I am pretty deeply entrenched in the Apple ecosystem, the recently-announced $50 Dot was so inexpensive I could not resist checking it out. (Before I go further: I work for Microsoft, so take that into account as you see fit.)

Out of the box, the Echo is very easy to setup for basic queries “Alexa, what’s my latitude and longitude?” and so forth. The Echo has a relatively lo-fi speaker and the integration with Sonos (what Amazon calls an “Alexa Skill”) is not yet available, so I haven’t used it all that much.

But there’s an API so you know I had to program something. My preferred solution for “computations in the cloud” is definitely Azure Functions written in F#, but for my first Alexa Skill I used Amazon Lambda running Python.

The first thing to focus on is that Alexa Skills are a separate service that can be programmed many ways, so there’s always going to be a certain amount of integration overhead in the form of multiple tabs open, jumping back and forth between the Alexa Skills and the Web server/service where you are handling the computation.

The Alexa Skills documentation is good, but there’s a good number of parts and I think it’s wise to write your first skill using Amazon Lambda, as I did. Amazon Lambda is often the default service in the documentation and there are often hyperlinks to the Lambda-specific page to do “X.”

A Skill for Gravity

A friend was talking to me about riflery and astonishing me with the flight times he was talking about. Alexa failed to answer some basic questions about ballistics (Alexa seems to me less capable than Google Assistant, Cortana, or Siri at answering freeform questions), offering me the perfect simple use-case for my first skill.

Minimum viable query: "What is the speed of an object that has fallen for 1.5 seconds?"

SWAG achievable: "How long would it take for an object dropped from the height of the Empire State Building to fall to the ground on Mars?"

The nice thing about my minimal query is that it’s both stateless and easy to answer with some math: all you need to answer is the duration of the drop and use a gravitational constant of -9.81. (Conversions from meters/second can come later.)

I followed the documentation on building an Alexa skill with a Lambda function to create an Alexa Skill named called “Gravity.” After naming, the next page of the Skill development site is “Interaction Model.” This is where I was shocked to discover:

Alexa doesn’t do natural language processing!

I ASS-U-ME’d that I would be receiving some programmatic structure that told me the “nominal subject” of the sentence was the noun speed and would allow me to search for a “prepositional modifier” whose “object” was the noun seconds and extract its modifier. That would allow me to recognize either of these sentences:

  • What is the speed of an object that has fallen for 1.5 seconds?; or
  • What's the velocity of an apple after after 1.5 seconds?

Or any of a large number of other sentences. Foxtype will show you such parsing in action at this (fascinating) page.

But no! As you can see in the screenshot below, the mapping of a recognized sentence to a programmatic “intent” is nothing but a string template! You either have to anticipate every single supported structure or you have to use wildcards and roll your own. (Honestly, I imagine that it’s not a long road before the wisest interaction model is Parse {utterance}.)


To be clear: ‘just’ voice recognition is extraordinarily hard and doing it in ambient environmental noise is insane. It’s only because Alexa already does this very, very hard task that it’s surprising to me that they don’t provide for some amount of the (also hard) task of parsing. The upside, of course, is that sound->utterance is decoupled from utterance->sentence. As far as I know, no one today provides “NLP as a Service” but it’s easy to imagine. (Although latency… Nope, nope, staying on topic…)

Returning to the screenshot above, you can see that it contains the bracketed template {duration}. The matching value will be associated with the key duration in calls to the Lambda function. And, to be honest, it’s a place where Alexa Kit does do some NLP.

You can help Alexa by specifying the type of the variables in your template text. For instance, I specified the duration variable as a NUMBER. Alexa does use NLP to transform the utterances meaningfully — so “one and a half” becomes “1.5” and so forth. I haven’t really explored the extent of this — does it turn “the Tuesday after New Year’s Day” into a well-formed date and so forth?

Alexa packages session data relating to an ongoing conversation and intent data and performs an RPC-like call (I actually don’t know the details) to the endpoint of your choice. In the case of Amazon Lambda, that’s the Amazon Resource Name (ARN) of your function.

The data structures it passes look like this:

  "session": {
    "sessionId": "SessionId.07dc1151-eb4e-4e12-98fa-64af3f59d82a",
    "application": {
      "applicationId": "amzn1.ask.skill.443f7cb5-ETC-dbecb288ff2d"
    "attributes": {},
    "user": {
      "userId": "amzn1.ask.account.ETC"
    "new": true
  "request": {
    "type": "IntentRequest",
    "requestId": "EdwRequestId.13cf7a2b-0789-4244-879f-f4fae08f315f",
    "locale": "en-US",
    "timestamp": "2016-11-18T17:24:09Z",
    "intent": {
      "name": "FallingSpeedIntent",
      "slots": {
        "duration": {
          "name": "duration",
          "value": "1.5"
  "version": "1.0"

The values in the session object relate to a conversation and the values in the request object belong to a specific intent — in this case the FallingSpeedIntent with the duration argument set to “1.5”.

On the Lambda side of things

Amazon Lambda has a template function called ColorIs that provides an easy starting point. It supports session data, which my Gravity skill doesn’t require, so I actually ended up mostly deleting code (always my favorite thing). Given the JSON above, here’s how I route the request to a specific function:

def on_intent(intent_request, session):
""" Called when the user specifies an intent for this skill """

print("on_intent requestId=" + intent_request['requestId'] +
", sessionId=" + session['sessionId'])

intent = intent_request['intent']
intent_name = intent_request['intent']['name']

# Dispatch to your skill's intent handlers
if intent_name == "FallingSpeedIntent" :
return get_falling_speed(intent, session)

def get_falling_speed(intent, session):
session_attributes = {}
reprompt_text = None
should_end_session = True

g = -9.82 #meters per second squared

if "duration" in intent['slots']:
duration = float(intent['slots']['duration']['value'])
velocity = g * duration**2

speech_output = "At the end of " + str(duration) + " seconds, an object will be falling at " + ('%.1f' % velocity) + " meters per second. " + \
speech_output = "Pretty fast I guess."

return build_response(session_attributes, build_speechlet_response(
intent['name'], speech_output, reprompt_text, should_end_session))

(Boilerplate not shown)

My Westworld prediction

var k = Convert.FromBase64String("vlqnRQo8YYXdqt3c7CahDninF6MgvRnqNEU+/tcbWdM=");
var iv = Convert.FromBase64String("gaXwv734Tu3+Jw1hgtNrzw==");
DecryptStringFromBytes(Convert.FromBase64String("Yr2XWzCxceStAF1BaUgaqmWcqFjzWskDDN4foaxfGEO5JHc/oKvgukkMHZuOiw+dK0JxnOhzC1ZA3QLqZZsQxFtjX+qvu0VRM0p6VEfcv18="), k, iv);

Keras is the Deep Learning Toolkit You Have Been Waiting For

I remember a time when “learning about computers” invariably started with the phrase “computers only operate on 0s and 1s…” Things could vary a little for a few minutes, but then you’d get to the meat of things: Boolean logic. “All computer programs are formed from these ‘logic gates’…”

I remember a poster that illustrated Boolean logic in terms of punching. A circuit consisted of a bunch of mechanical fists, an “AND” gate propagated the punch when both its input were punched, an “OR” required only one input punch, etc. At the bottom were some complex circuits and the ominous question: “Are you going to be punched?” Because Boston. (The answer was “Yes. You are going to be punched.”)

Anyway, the point is that while there was a fundamental truth to what I was being told, it was not overwhelmingly relevant to the opportunities that were blossoming, back then at the dawn of the personal computer revolution. Yes, it’s important to eventually understand gates and circuits and transistors and yes, there’s a truth that “this is all computers do,” but that understanding was not immediately necessary to get cool results, such as endlessly printing “Help, I am caught in a program loop!” or playing Nim or Hammurabi. Those things required simply typing in a page or two of BASIC code.

Transcription being what it is, you’d make mistakes and curiosity being what it is, you’d mess around to see what you could alter to customize the game, and then your ambition would slowly grow and only then would you start to benefit from understanding the foundations on which you were building.

Which brings us to deep learning.

You have undoubtedly noticed the rising tide of AI-related news involving “deep neural nets.” Speech synthesis, Deep Dream’s hallucinogenic dog-slugs, and perhaps most impressively AlphaGo’s success against the 9-dan Lee Sedol. Unlike robotics and autonomous vehicles and the like, this is purely software-based: this is our territory.

But “learning about deep learning” invariably starts with phrases involving the phrases “regression,” “linearly inseparable,” and “gradient descent.” It gets math-y pretty quickly.

Now, just as “it’s all just 0s and 1s” is both true but not immediately necessary, “it’s all just weights and transfer functions,” is something for which_eventually_ you will want to have an intuition. But the breakthroughs in recent years have not come about so much because of advances at this foundational level, but rather from a dramatic increase in sophistication about how neural networks are “shaped.”

Not long ago, the most common structure for an artificial neural network was an input layer with a number of neural “nodes” equal to the number of inputs, an output layer with a node per output value, and a single intermediate layer. The “deep” in “deep learning” is nothing more than networks that have more than a single intermediate layer!

Another major area of advancement is approaches that are more complex than “an input node equal to the number of inputs.” Recurrence, convolution, attention… all of these terms relate to this idea of the “shape” of the neural net and the manner in which inputs and intermediate terms are handled.

… snip descent into rabbit-hole …

The Keras library allows you to work at this higher level of abstraction, while running on top of either Theano or TensorFlow, lower-level libraries that provide high-performance implementations of the math-y stuff. This is a Keras description of a neural network that can solve the XOR logic gate. (“You will get punched if one, but not both of the input faces gets punched.”)

import numpy as np
from keras.models import Sequential
from keras.layers.core import Activation, Dense
from keras.optimizers import SGD

X = np.zeros((4, 2), dtype='uint8')
y = np.zeros(4, dtype='uint8')

X[0] = [0, 0]
y[0] = 0
X[1] = [0, 1]
y[1] = 1
X[2] = [1, 0]
y[2] = 1
X[3] = [1, 1]
y[3] = 0

model = Sequential()
model.add(Dense(2, input_dim=2))

sgd = SGD(lr=0.1, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(loss='mean_squared_error', optimizer=sgd, class_mode="binary")

history =, y, nb_epoch=10000, batch_size=4, show_accuracy=True, verbose=0)

print (model.predict(X))

I’m not claiming that this should be crystal clear to a newcomer, but I do contend that it’s pretty dang approachable. If you wanted to produce a different logic gate, you could certainly figure out what lines to change. If someone told you “The ReLu activation function is used more often than sigmoid nowadays,” your most likely ‘let me see if this works’ would, in fact, work (as long as you guessed you should stick with lowercase).

For historical reasons, solving XOR is pretty much the “Hello, World!” of neural nets. It can be done with relatively little code in any neural network library and can be done in a few dozen lines of mainstream programming languages (my first published article was a neural network in about 100 lines of C++. That was… a long time ago…).

But Keras is not at all restricted to toy problems. Not at all. Check this out. Or this. Keras provides the appropriate abstraction level for everything from introductory to research-level explorations.

Now, is it necessary for workaday developers to become familiar with deep learning? I think the honest answer to that is “not yet.” There’s still a very large gap between “what neural nets do well” and “what use-cases are the average developer being asked to addressed?”

But I think that may change in a surprisingly short amount of time. In broad terms, what artificial neural nets do is recognize patterns in noisy signals. If you have a super-clean signal, traditional programming with those binary gates works. More importantly, lots of problems don’t seem easily cast into “recognizing a pattern in a signal.” But part of what’s happening in the field of deep learning is very rapid development of techniques and patterns for re-casting problems in just this way. So-called “sequence-to-sequence” problems such as language translation are beginning to rapidly fall to the surprisingly effective techniques of deep learning.

… snip descent into rabbit-hole …

Lots of problems and sub-problems can be described in terms of “sequence-to-sequence.” The synergy between memory, attention, and sequence-to-sequence — all areas of rapid advancement — is tipping-point stuff. This is the stuff of which symbolic processing is made. When that happens, we’re talking about real “artificial intelligence.” Artifical intelligence, yes, but not, I think, human-level cognition. I strongly suspect that human-level, general-purpose AI will have a trajectory similar to medicine based on genetics: more complex and messy and tangled to be cracked with a single breakthrough.

Debugging provisioning profiles on the command line

Raise your hand if you’ve ever struggled with getting your app’s bundle identifier, info.plist, and entitlements.plist to match up with your provisioning profile.

I tried to explain provisioning profiles using the ten-hundred most common words, but in slightly-less-common words, a development prov-pro associates: A team, a developer, an application identifier, privacy and security entitlements, and development devices.

While there’s no silver bullet, there is a way to dump the contents of a provisioning profile into a readable plist format. From the command-line, run:

security cms -D -i some.mobileprovision

Here, for instance, is the output of a provisioning profile for an app that uses SiriKit to trigger a workout:


As you can see, this is a convenient way to confirm the associations in the prov-pro, particularly entitlements, the app ID, and provisioned devices.

Mysterious crashes in your iOS 10 program? Check your info.plist

If you’re developing for iOS 10 and your app “silently” crashes (especially if it’s an older app), the culprit could well be the increased privacy requirements in iOS 10. Namepaces such as HomeKit now require specific privacy-related keys to be in your info.plist (for instance, NSHomeKitUsageDescription). If you don’t have them, the system automatically closes your application without an exception or Console.log message (if you run in the simulator, you may see a PRIVACY_VIOLATION notice in the stack trace).

Streaming a Web video to AppleTV with Xamarin

If you have the URL of a streaming video, it’s easy to display on an AppleTV, even though tvOS does not have a UIWebView (which would make it really easy). You have to use some AVFoundation code, such as:

var src = NSUrl.FromString("https://somevideo");
var asset = AVAsset.FromUrl(src);
var playerItem = new AVPlayerItem(asset);
var player = new AVPlayer (playerItem);
var playerLayer = AVPlayerLayer.FromPlayer (player);
//Might want to modify this so that it's the same size as the source video
var frame = new CGRect (0, 0, this.View.Frame.Width, this.View.Frame.Height);
playerLayer.Frame = frame;
this.View.Layer.AddSublayer (playerLayer);
player.Play ();

Note: This won’t work with normal YouTube page URLs since the YouTube stream URLs are not directly accessible.

The Half-Baked Neural Net APIs of iOS 10

iOS 10 contains 2 sets of APIs relating to Artificial Neural Nets and Deep Learning, aka The New New Thing. Unfortunately, both APIs are bizarrely incomplete: they allow you to specify the topology of the neural net, but have no facility for training.

I say this is “bizarre” for two reasons:

  • Topology and the results of training are inextricably linked; and
  • Topology is static

The training of a neural net is, ultimately, just setting the weighting factors for the elements in the network topology: for every connection in the network, you have some weighting factor. A network topology without weights is useless. A training process results in weights for that specific topology.

Topologies are static: neural nets do not modify their topologies at runtime. (Topologies are not generally modified even during training: instead, generally the experimenter uses their intuition to create a topology that they then train.) The topology of a neural net ought to be declarative and probably ought to be loaded from a configuration file, along with the weights that result from training.

When I first saw the iOS 10 APIs, I thought it was possible that Apple was going to reveal a high-level tool for defining and training ANNs: something like Quartz Composer, but for Neural Networks. Or, perhaps, some kind of iCloud-based service for doing the training. But instead, at the sessions at WWDC they said that the model was to develop and train your networks in something like Theanos and then use the APIs.

This is how it works:

  • Do all of your development using some set of tools not from Apple, but make sure that your results are restricted to the runtime capabilities of the Apple neural APIs.
  • When you’re done, you’ll have two things: a network graph and weights for each connection in that graph
  • In your code, use the Apple neural APIs to recreate the network graph.
  • As a resource (download or load from file) the weights.
  • Back in your code, stitch together the weights and the graph. One mistake and you’re toast. If you discover a new, more efficient, topology, you’ll have to change your binary.

This is my prediction: Anyone who uses these APIs is going to instantly write a higher-level API that combines the definition of the topology with the setting of the weights. I mean: Duh.

Now, to be fair, you could implement your own training algorithm on the device and modify the weights of a pre-existing neural network based on device-specific results. Which makes sense if you’re Apple and want to do as much of the Siri / Image recognition / Voice recognition heavy lifting on the device as possible but allow for a certain amount of runtime flexibility. That is, you do the vast majority of the training during development, download the very complex topology and weight resources, but allow the device to modify the weights by a few percent. But even in that case, either your topology stays static or you build it based on a declarative configuration file, which means that whichever route you choose, you’re still talking about a half-baked API.


Review: 11-Day Diving on the Galapagos Master

Trip Review: Galapagos Master, 11-Night Liveaboard Diving

My wife and I recently returned from 11 days on The Galapagos Master, a 16-passenger liveaboard vessel whose itinerary includes Wolf and Darwin Islands.

The first thing to say about Galapagos diving is… Well, okay, the first thing to say about Galapagos diving is how incredible the fish life is. More on that in a minute…

The second thing to say about Galapagos diving is to talk about the temperature: temperature descriptions generally say something like “60-76F” and you might think “Well, I’ll plan for the middle of that estimate: 68F.” But that’s not right: the diving here is 60F or 76F, depending on where you dive. And even though almost exactly half the dives are in water that was in the mid-70s, the “feel” of the water temperature was determined by those in the 60F area. So 7mm hooded wetsuits and I envied the one person on our boat who dove in a drysuit. (My wife says her 5mm with a 3mm hooded steamer and a LavaCore was also okay, and she had more flexibility on the warmer dives.)

The other thing, for me, is gloves. I never wear gloves since in general I have no need to touch the reef. But in the Galapagos the large majority of dives involve tucking in to rocks and holding on in strong currents. Additionally, at Darwin, Wolf, and Cabo Douglas (Fernandina) there was surge.

And the rocks are covered in barnacles. I didn’t wear gloves for the first several dives and my hands got sliced up.

Dives are limited to 50 minutes. We were all diving nitrox and spending the majority of our dives at 60-80’, so I thought that duration was good: long enough to linger when the sights were good, short enough so that air consumption wasn’t a limiting factor, and brief enough that no-deco was very manageable (I had a few 4th dives where I was deco-limited.) In the cold water at Cape Douglas (marine iguanas) and Punta Vicente Roca (molas and penguins) the dives were shorter — 40 minutes. (The water was cold, but with the marine iguanas my max depth was 19’! I would have happily spent more time with them.)

Each buddy-team was given a DiveRite audible alarm powered by their low-pressure BCD inflator and a Nautilus Lifeline GPS/VHS radio. We never had any cause to use either, but the Nautilus, in particular, struck me as a showing a good concern for safety.

Itinerary and Diving

Our 11-day itinerary was: board the boat in Puerto Baquerizo Moreno on San Cristobal. Same day we had a 20-minute check-out dive in the harbor: cold, poor vis, just a chance to get your weight correct. (Some sea lions came in and played with us, which was fun.)

2nd day was a dive apiece at Mosquera and North Seymour Island. I thought these would be more “check out” style, but actually the dive at Mosquera was excellent! First hammerhead of the trip, a big school of mobula, schools of barracuda, steel pompano, big blue-spotted trevally, “some kind of bonito.” Our 2nd dive, at North Seymour, was apparently more-commonly a highlight, but we got somewhat skunked.

We did a brief land excursion on North Seymour and, for me, it was one of the highlights of the trip — our only chance to see nesting blue-footed boobies and frigate birds. We saw several males displaying, a few pairs “dancing,” and even one sitting on an egg. (I’m kicking myself at missing a post-dive-trip day-trip to Isla Lobos from San Cristobal to see more breeding birds.)

Then we steamed for Darwin Island. We apparently got a late start for this, leaving North Seymour at 4:30PM when we were “expected” to get out of there at 1:30PM. (But that’s a little confusing to me, as we may have dawdled an extra 30 minutes or whatever on the land expedition, but the overall schedule was set by the boat.)

The upshot was that we didn’t arrive at Darwin until 8:30AM and dove immediately. The next 4 days (2 at Darwin, 2 at Wolf, 4 dives per day) were amazing. Warm, with occasional hints of a thermocline, moderate-strong currents (I think once we had a spot with something like 2 knots), insane density of fish. Jacks, hammerheads, Galapagos sharks, yellowfin tuna, smaller tuna… just amazing.

Visibility was not “murky” but it was definitely “hazy.” Maybe 25-35’-ish total, but things at the limits of visibility were definitely more silhouette-y than defined. So even though there was tons of wildlife, you would really only very-clearly see maybe 3-4 close passes per dive.

Our panga (“Jaguar Sharks!”) was quite experienced (professional divemasters, marine biologist, etc.: with just over a thousand dives, I was by far the least experienced). I think on a difficulty scale of 1-10 for recreational diving, this was 7-8 stuff: cold, currents and surge, limited viz. This would not be a place for divers uncomfortable with their gear.

Additionally, particularly at the south side of Darwin’s Arch, if you drifted away from the group and were not in the panga right around that 50-minute mark, you could get close to some extremely dangerous wave breaks. The dive guides knew the topography and drifts very well and if you paid attention to the rules and stuck with them, it was all fine. But again, it was the type of place where a mistake that separated you from the group could get very serious, very quickly.

I could go on for thousands of words detailing the diving, but suffice it to say that it was great. There were endless schools of predators such as jacks and bonito as well as reef fish such as creole fish. The sharks varied depending on the specific dive location and time of day, but typically you’d settle in at 3 or 4 stops along the reefs and usually when you settled in, some amount of hammerheads and Galapagos would come by. Sometimes, when the current was strong, you’d be in a perfect situation where the hammerheads were slowly making their way up-current and it was just an unending conveyor belt.

Another fun thing to do at Darwin was to drift past the boat on its mooring: there must have been 20 silky sharks swimming along in its eddy and if you could hold on to a panga line or get into the eddy at the stern of the ship, you would just be surrounded by silkies. The islands and birdlife of Darwin and Wolf were fascinating to observe with binoculars from the stern of the ship.

After Darwin & Wolf, the diving was one location per day, usually with a single dive site dived only 3 times per day. In our case, we often dove, 7:45AM, 9:45AM, 11:45AM. Generally diving did seem to deteriorate as the morning progressed, so the only way I’d change that schedule would be a dive before breakfast, but that was never presented as an option to us. I think there was one more day when we had an after-lunch dive.

Fernandina had a beautiful deep-dive to see horn sharks and red-lipped batfish (coldest dive, with 58F on my computer and 95’ of neoprene compression). After that, we did 2 dives in 5 meters to see marine iguanas feeding. Absolutely amazing. I do want to say that when we first got in, I experienced the most powerful surge of the trip: the surge was so strong that it caused my octopus to freeflow and then, even with a good grip, I got peeled off a rock and pushed a solid 10 yards. Again, this is a situation where a less experienced person could make a serious mistake and try to re-grip rather than accept spending the next few waves being washed back and forth.

Another highlight of the trip then occurred: while crossing from Fernandina to Isabella I spotted a pod of orcas in the distance. They approached the boat and checked us out for ten minutes or more, swimming right alongside the boat, tilting on their sides to look up at us, etc.

As with the iguanas at Fernandina, the next two days were destination dives as well: one day to see molas (ocean sunfish) and penguins (snorkeled with one) and the next to see pelagic manta rays. These were fine and again the walls were beautiful, with abundant black coral and bushy brown gorgonians teeming with long-nose hawkfish.

Then a long cruise to Cousin’s Rock near Santiago and the final 2 dives of the trip. I feel silly downplaying any diving that involves a cave filled with white-tipped reef sharks and sea lions but compared to the other diving on the trip, this was anti-climactic.

The final half-day of the trip involved a bus ride to a farm in the Santa Cruz highlands to see giant tortoises. This was quite good: they were free-ranging and it seemed more natural than seeing them in pens. As a reminder of the threat of introduced species, I was bitten by a fire ant as I watched a giant tortoise.

Then we went down to Puerto Ayora and spent several hours, having a couple drinks and lazily shopping for souvenirs. The bus picked us up at 6:30PM and got us back to the boat near 8PM, where we had a final cocktail reception where the “tipping” occurred (more on that below) and then dinner. (Again, this was a case where the schedule was set by the boat, so the fact that we were all starving by the time we were fed near 9PM seems like something they could adjust.)

The next day we were back in Puerto Baquerizo Moreno and taken off the boat at 830AM.

The [official itinerary] describes the cadence as an early dive, breakfast, morning dive, lunch, and two afternoon dives. That’s not at all how ours worked: first, we never had an opportunity to dive before breakfast and on 3-day dives (most days not on Darwin and Wolf) we often did all 3 dives before lunch with only 45-minute breaks. That was fine, since diving was generally better in the morning and getting in and out of a cold wetsuit is a pain.


The Galapagos Master is the former Deep Blue (so you might search for other reviews under that name). There are 4 below- and 4 above-deck cabins. We had an above-deck cabin but I do not think they were worth extra: the windows did not open and being that far above the center-of-gravity of the ship may have made the motion a little more obvious. Other than the standard shipboard reaction of “Oh my god, where are we going to put all our stuff?” the two things that stand out are the beds, which were very uncomfortable (pads over wood, with two single beds pushed together so that the rails created a “chastity bump”), and the electric toilet, which was absolutely incapable of clearing solids (if you know what I mean) with anything less than 4-5 flushes. Toilet paper goes into a container at the side.

There is a salon where the in-door socializing happens, with a big-screen TV with HDMI inputs, so it was easy to do slideshows or watch movies from computers. The mess had one large table and a few smaller ones. It was well-configured for socializing. The sundeck was the major socializing place, and once clearly sported a bar and chairs, but only one chair was attached. Instead, you just lounged along the rails.

I’m not a “foodie” and I thought the food was fine, but I think there was a little eye-rolling from some more refined people. There was always a salad and some amount of vegetables, and then usually a fish and a meat dish with some starch like potatoes or rice or plantains. Often meals started with a soup and there was always a dessert. There were two vegetarians on our trip and the galley seemed to be able to accommodate them well enough. Soft drinks were complimentary, beers were $3 apiece and cocktails and spirits were $6 apiece. Bottles of wine were $25.

The dive deck was quite good, with individual stations along the rails, a wetsuit cleaning-tank and rack in the middle, 2 hot-water hoses with shower nozzles, and a large cameras-only tank. Under the tanks were cubby-holes with milk crates in which you kept your miscellaneous dive gear. Up a few steps from the dive deck was a passage with a long camera bench with 2 air blowers (well, 3, but one wasn’t working). On the other side there would be post-dive snacks and hot chocolate or ice tea. A nice thing were post-dive towels, neatly labeled with your station number, so you would be assured of getting one.

On the first day our nitrox was a little low, at 29+, but mostly we dove around 32% O2. Again, I thought the nitrox vs. time vs. depth balance was just right: you ended the 4-dive days close to deco-limited but I never got close to depth-limited.


This is a pet peeve of mine. I’m from the US and I tip well because I know that, when “tipping” is a big part of how workers make their money, workers get absolutely screwed. Our trip had people from the US, England, and Germany, all of which have vastly different attitudes and expectations about tipping. And although “it’s absolutely up to you” there is a “recommended 10% tip” on your $600-a-day-per-person dive trip that is “an important part of how the crew make money.” This is total BS! If a fair wage for the crew amounts to $60 per day per guest, charge $660 per day and make tipping truly optional.

As one guest from England, who was not prepared to tip in cash (which is the only way), said “Half the trip fee goes to the cost of fuel for the ship. Why am I paying 10% of the fuel cost for ‘service’?”

Also, “tipping” this way creates perverse incentives. After safety, the most important role of the diveguides is to enforce the conservation rules, but that’s made more difficult when you rely on “tips” as a major part of your income for the trip. Which brings us to…

“That Guy”

One of the things we got ready to board the boat is that “there’s always one guy.” In our case, he was a German who fancied himself a “photo-journalist.” What that meant was that in his mind, because he made a few thousand Euros per year from stock photography fees, he was justified in breaking the rules: he didn’t stay in a line so that all divers were at an equal distance from the skittish hammerheads, he dive-bombed other photographers, he pushed my wife out of the way when she had the temerity to videotape a marine iguana with “just” a GoPro, and, worst of all, he would swim up to skittish animals such as hammerheads or mola and blast them with his strobes. He was clearly out of line time-and-time again, and the dive guides never confronted him.

This became a topic of every post-dive talk and we talked to our dive guide about it. He spoke at dinner about the importance of obeying the no-harassment rules and the dire consequences of breaking them. Then again, at breakfast the next day, he reiterated the importance of not chasing the animals.

And then, an hour later “That Guy” bombed a Mola and chased it away. Back on the panga, the other divers were saying stuff and when it became clear that the dive guide wasn’t going to say anything, I gave the guy both barrels. The upshot? Well, he didn’t dive with us anymore, but essentially he got a private dive guide and (according to reports) had a great time swimming up to and blasting pelagic mantas.

Such behavior will continue as long as the rest of us divers, whether photographers or not, tolerate it. We all want to see the animals as well as possible, we all paid a lot of money, we all would love a photo. But sometimes nature doesn’t accommodate our wishes. What we do in those circumstances is the test of our character and, if you label yourself a “photographer” (much less a “photo-journalist”), a test of your ethics.

Don’t be “that guy.”

WWDC Remote Viewing Protips

I attended the 2015 WWDC and made these notes afterwards. Aside from the specifics re. the Apple Watch and AppleTV, they may be of value to those who are considering streaming sessions next week:

WWDC: Post-show Streaming is the Key to Value

From an editorial perspective, one thing that is clear about WWDC is that the main audience for the sessions is not the developers in attendance, but the much more diverse, more diffuse, and more transient on-line audience that will view the videos over the next months and even years.

WWDC Session Videos are great as overviews, poor as references

What I’ve come to realize is that WWDC sessions are great as overviews, but poor for depth. They are very much worth watching when you’re new to a framework, they’re somewhat worth watching if you haven’t programmed in the framework lately (you might see some class you hadn’t appreciated), but they are not the place to discover a way out of some corner-case or programming limitation.

Microsoft explicitly labels the depth of their conference talks as being 100-, 200-, or 300-level, and 300-level content at WWDC was vanishingly rare. (As I write this, I can only speak to the talks I physically attended, but several talks definitely promised more depth than they delivered.)

I wonder if this is an artifact of The Dog That Didn’t Bark aka Apple TV. It must have been pulled very late. Both Xcode and Apple’s Developer Site, which had to be updated to support the new OS betas, are littered with Apple TV references. Perhaps it was the case that some of these talks were put together quickly. (Although you wouldn’t guess it from the universally well-practiced speakers.)

The real keynote was the Platform State of the Union

Monday’s keynote was covered by news vans and live blogs and all that crap. There was, perhaps, 5 minutes of developer content in this 2.5-hour stemwinder. From the audience, anyway, the music stuff was awkward to the point of embarrassment.

Skip it and watch [Platform State of the Union] instead. This was the true developer’s keynote and contains an excellent overview of El Capitan, iOS 9, and watchOS. (By the way, the witty kids pronounce “watchOS” so that it rhymes with “nachos.”)

The Shocking Secret You Can Use to Determine Which Videos to Stream

Is that a proper 21st century headline?

Anyway, here’s the key: many sessions followed a standard naming practice:

— “Introduction to…” talks are 100-level (if that) “tables of content.” They hardly have any code on screen, but contain references to other videos that provide the 200- or 300-level content. If you’ve ever programmed in the namespace before, you can skip these talks.

— “What’s New In…” talks are 100-level “Release Notes.” There may be some code, but what you’re really looking for here are the new classes and general new capabilities. This is the video with which you should start if you have programmed in the framework before, even if you’re pretty comfortable. Again, all of these talks are good at referencing other, more substantive, talks. This is my main recommended tactic for finding deep content on frameworks with which you are familiar: it’s much more effective than guessing from session titles and descriptions.

— Beware talks that have the words “tips”, “tricks,” or “practices.” These were the talks that disappointed me. Such words traditionally mean 300-level content. If you’re an attendee and you’re budgeting precious in-conference time to “tricks” and “practices,” that’s a strong indicator that you’re familiar with the framework and are encountering its limitations and corner cases. But at WWDC, these sessions appear to be more focused on the newcomer or relatively inexperienced framework user.