Risky Autonomy vs. Walled Gardens

Issue #12: A story about tensions, EigenLayer guest talk, and sci-fi contest extension

In this issue: Artist, researcher, and software developer Sarah Friend (SoP23) explores a tension central to the grey areas of high stakes software. Also – join us tomorrow for a guest talk with EigenLayer CEO, Sreeram Kannan, and read on for an update and deadline extension about the terminological twists contest, which has received 30 great entries so far!

My phone used to always go off in public. The tone was like a spaceship landing, I’m told—though it was just the default for my phone’s model, the OnePlus 12. I wonder what people thought when they heard it. Why was I getting phone calls all the time? Was I an on-call engineer at a very important (and obviously top secret) organization? Or a negligent mother, ignoring an emergency call from my childrens' school? They might wonder why I didn’t fit the communication protocols of my generation: text please, voice calls only when scheduled. Maybe I was in fact 70 years old with an off the hinge skincare routine? No. None of these things.

One time, I was at a performance at Gropius Bau, a large contemporary art gallery in Berlin. White walls, vaulted ceilings—definitely the kind of thing you’re supposed to have your phone on silent for. It was a satirical cooking show, and the host had been pacing through the crowd to get to the different cooking stations, arranged around the perimeter. He was right beside me when the spaceship started descending. I jumped, fumbling my phone out of my bag and tried to silence it as quickly as possible. Never fast enough.

He looked down at me, my hand still guilty on the volume button. “Uhh, thank you” he sneered over the microphone for the whole room.

Heart-pounding, I decided to stick up for myself. I imitated his fake smile to the best of my ability and batted my eyelashes, “It’s a medical alarm.”

He was horrified, of course, but he didn’t have time to ask questions. He hissed an apology off-microphone and carried on with the show. I took some candy out of my purse and popped it in my mouth.

I wasn’t lying. You see, I’m an insulin-dependent diabetic. And the alarm couldn’t be turned off.

All people with Type 1 diabetes (what I have, an autoimmune condition) and some people with Type 2 diabetes (more common, a metabolic disorder) use insulin and sugary foods throughout the day to keep their blood sugar within a safe range. Their body can no longer do so on its own. High blood sugar is bad for you, it can cause a lot of problems, but only in the longer term. Low blood sugar is a medical emergency, as in it can kill you. It can happen quickly; not hours but minutes. Thus the alarm.

Being diabetic is many things—it's stressful, it’s tiring, it provokes uncomfortable existential questions. Another thing it means is that the body is intimately integrated with software. There are a variety of devices to aid the dosing of insulin and measuring of blood sugar, many software-mediated, including insulin pumps and continuous glucose monitors. I have been using the latter, and it’s the device that was responsible for my pet spaceship: the 24/7 ringer that got me angry stares in the movie theatre, confused glances in the grocery store checkout line, and woke me up at midnight. The cgm streams my blood sugar data to my phone in real time. But this, my greatest tool in diabetic management, began over time to also feel like the villain in a psychological drama. It pursued me ceaselessly, watching round the clock: a sonic assailant that I could never escape because it of course was part of myself … until it wasn’t.

Browsing reddit one day after a hardware upgrade, I found a post on reddit leading me to a github repo for something called “xDrip”, an open source android app for reading cgm signals. I downloaded the .apk and installed it, expecting trouble. To my surprise it worked perfectly: I found a switch to disable all alarms for as long as I wanted, the ability to delete the built-in alarms entirely and add my own custom ones, triggered at whatever blood sugar levels I chose. I could add a different profile for different times of day, enabling less aggressive numbers at night (and therefore more nights of uninterrupted sleep), and even a different sound for each alarm. Just like that, I had root access to the spaceship. I knew I had been in something like a walled garden—a closed system where the possibilities had been artificially constrained. What I hadn’t realized is just how easy it would be to walk out.

Are apps like this legal? Maybe, depending on your jurisdiction. Where I live, my doctor is not allowed to recommend or even mention it, but she’s not legally obligated to stop me from using it if I choose. Is it safe? For me, it has improved health outcomes—physical and mental, but I won’t deny that for some patients it may be more dangerous. Not everyone is as comfortable with open source software as I am. And not everyone with diabetes is an adult. It’s easy to understand why a manufacturer would make the most urgent category of alarms unsilenceable. But I’ve chosen my risky autonomy, and you can take it from my cold dead hands - a phrase I don’t use only idiomatically, since my life is in fact a chip on the table.

In Chiang Mai, at Edge Lanna, fresh in my second month of open-source freedom, I started to reflect on how what I’d experienced with cgm apps spoke to a more widespread protocol tension. One day, I was walking to lunch—with some difficulty, as crossing the street in Chiang Mai is an art. In Germany, where I live, people will wait for a red crossing signal to turn green even in the dead of night on an empty road. Old ladies will scold you for jaywalking. This same behaviour in Chiang Mai will have you waiting on the same side of the road you started on, long after your hunger has faded into the clear-eyed alertness of fasting and your lunch date has given up on you and gone home.

There is a lot of traffic, and far fewer crosswalks and streetlights than you might expect. You must instead watch traffic closely and seize any opportunity you see. Besides cars, Thailand has the highest motorcycle usage in the world. Scooters and motorcycles swarm between lanes. Google Maps has adapted, offering different routes and travel times for scooters vs cars. There’s no doubt that being able to squeeze between cars can get you to your destination faster, and indeed the scooter travel estimate is often the fastest one on display, but it’s definitely a trade off that comes with some risk: Thailand notably ranks highest in motorcycle fatalities, as well. The motorcyclists have chosen risky autonomy. The drivers of cars perhaps less so. But everyone on the road in Thailand is experiencing more risky autonomy than the walled garden of German crosswalks. I contrasted both to the even stricter system of air travel that had brought me to Thailand just a few days before, with its literal security perimeter.

During one of our workshops, I presented the dichotomy of “risky autonomy” vs “walled garden”. Our group found examples everywhere: Linux vs Apple, a tradfi bank vs crypto, different political systems. Some enclosures are opt-in (e.g., Apple) others are not (traffic laws). One core thing that matters, as I think with all assessments of risky autonomy vs walled garden, is which direction the risk is pointing. Risk is not a scalar, it is a vector, which is to say it’s not experienced uniformly across a system nor does it have the same pattern in all systems. When I, an adult with no dependents, bypass official software for diabetic tech, the risk is substantially towards myself. With traffic, this is less clear and changes based on what kind of vehicle you’re driving (scooter, car, plane)—the likelihood of each of these to harm people who aren’t the driver (or pilot) is very different—and that’s likely why, in both Chiang Mai and Germany, drivers of cars experience and consent to more regulation than pedestrians.

But of course, this framing of walled garden/risky autonomy as two poles in opposition might itself be a walled garden. Is the "dichotomy" itself another artificial constraint? You can ask where a specific system sits between two forces, or how the choice between them is made, but the conversation really gets interesting when you start to unravel the dichotomy itself. For example, thinking about the direction of risk in a system reveals key differences between cgm app features and traffic laws. Risky autonomy and walled garden are present as forces in both systems, but there's more on the table. And in some cases, if we focus on this specific "more" we may be able to imagine our way out of a trade off entirely. In order to enable developer mode on Android, you have to go to Settings -> About phone and tap the Build number 7 times, like entering a magic password. This action is presumed to be so unusual that anyone undertaking it must know what they're doing, and any corresponding increase in the danger is assumed to be understood. With another open source diabetic tech project, Android APS, a new user has to use the app with a series of levels, which take a minimum duration of 28 days to be completed. When the diabetic user successfully completes all objectives, like playing through levels in a video game, they receive full access to all features. A pattern like this could go in cgm apps by default. Or if you want an appeal to authority, we can imagine an auth code that can only be input by an endocrinologist. Or a signed legal waiver. Or maybe something else. By more closely considering benefits for each type of system, we can at least start to suggest ways to resolve the tension: to move beyond picking a side, towards the optimal.

---

Tomorrow’s Guest Talk

The founder and the CEO of EigenLayer, Sreeram Kannan, will join us tomorrow, April 16th, at 10:00 PDT (18:00 UTC) to talk about technical protocols, different kinds of trust, and innovation in conversation with Venkatesh Rao. Click on the YouTube thumbnail to find the “notify me” button, and receive a one-time reminder notification.

---

Terminological Twists Update and Deadline Extension

We have extended the deadline for our premiere writing contest, Terminological Twists, until Monday, April 21st. As of today, 30 submissions have come in so far. You can keep track of the competition on the contest page, where we’ll post an updated entry count, daily, until the submission period closes.

Terminological Twists Challenge

Shortlisted stories could be published as soon as next week. We’re excited about the entries so far and have been impressed by the creative use of LLMs in the writing and editing process.

If you have questions about the contest, please direct them to the #scifi-contest channel on our Discord server. The Discord is also where we process pitches to contribute to Protocolized and is a great place to get an early scoop on the latest in protocol studies.

---

SoP25

Thank you to everyone who applied to this year’s Summer of Protocols curriculum development program. The interview process has concluded, and we look forward to sharing more news soon.

Last week’s issue:

The Signal Under Innsmouth