Tenko Logo

Back in 2020, after having spent the best of the past two decades in bioinformatics (first in academia, then in industry), tinkering with machine learning and data science to cure cancer™, I decided it was time to move onto the next challenge: solving climate change™.

Ideation phase took a while, and wasn’t as smooth as planned, due to some, err, unforeseen circumstances. Things went a bit like:

  1. First half of 2020: getting asked about 5 times a week to join in on some new medtech/pharmatech venture “planning for the next pandemic”. 😑
  2. Summer 2020: “hmm, I wonder if that whole climate thing might not be the next big thing we need to worry about. Perhaps this might just be the right intersection of personally care about, topical to my skillset, and might get paid for. Dare I say ikigai!?”
  3. Late 2020: exploring the space and spending endless hours on video calls with everyone I know somehow involved in climate and climate tech. Realising that:
    • Consumer advocacy apps are great, but not my jam.
    • Carbon counting is a bit of a scam <cough>greenwashing</cough>, and by 2023, there will be more carbon counting apps/websites/services than humans on earth.
    • Mitigation (carbon capture and other cool hardware moonshots) is great, but my physics/mechanical engineering days are a little too far in my past…
    • Everyone is so busy putting little Carbon Neutral by 2050 stickers on their website and recycling coffee pods at the office, that literally no one has a real clue how climate change is impacting us now. And very clearly it is.
  4. New Year’s Eve 2020: settling on the decidedly unsexy but intriguing challenge of modelling the economic impacts of climate change. On business (a girl’s gotta eat). The real fun begins: which impacts (weather disasters? geopolitical strifes? taxes? regulation?), which businesses (mustard seed growers? semiconductor manufacturers? logistics companies?), which regions (South East Asia? Europe? North America?)… Deciding to keep it simple and do it all.

And here I am, 18 months later, working with a team of brilliant people, building some very cool models and data products centred around climate, risk, and economics.

My personal outlook on the climate crisis is considerably more bearish than the IPCC consensus (itself a lot darker than the mainstream view). In the immortal words of my old friend John Maynard: in the long run, we’re all fucked. In the fairly short run: also.

But one of my few bright hopeful spots in that bleak landscape, is the potential for existing technologies to assist us with resilience: remote sensing data that can predict in near-real-time when and where crops will fail, roads will flood, or factories will shut down, and help us find the best way to minimise these impacts. We will need that level of scale and precision, to measure and adequately respond to the global crisis currently unfolding.

As I often do when having to name a new project, I lazily dug into Japanese for this one.

Kikō (気候), the word for climate, was a little too generic for trademark purposes, but Tenkō (天候) a concept halfway between climate and weather (天気, tenki) had both a nice ring to it, and some auspicious homonyms (転向: change of direction…).

As it turns out, the choice was fitting: Tenko (the company) is not really about climate or climate change (this vague and distant threat that people keep pushing to some distant horizon in 20 or 30 years). Rather, it is about climate-related impacts (and particularly the weather, but not only).

More generally, Tenko is about using data (of which there is a lot) to measure risks (of which there are a lot).


There’s nothing new in whining about tech monopolies: the companies that enjoy them and the doom that awaits us for foolishly trusting them.

At this point, we have all been at either end of a speech on the dangers of letting Microsoft/Google/Facebook’s dominion over our life go unchecked and unbalanced. A speech that usually ends with one party’s eyes glazing over and excusing themselves to the restroom.

The problem with standard denunciations of these potential abuses is that they tend to rest on abstract, distant and mostly theoretical arguments. It’s not that we don’t care about fair competitive practices and healthy markets, it’s just that we care a lot more about convenience and not messing with things that run kinda-ok. If the Galactic Empire could ensure that the Alderaan-Tattooin express shuttle runs on time, most of us would be fine with their hegemonic market practices, and tell that troublemaker Luke Skywalker to go seek therapy for his Freudian issues instead of blowing things up.

But the reality is that, if you use any of the services provided by these monopolistic behemoths (and even if you don’t), there is a statistical certainty that it will bite you in the arse at some point. And when it does, that monopolistic behemoth status means you will be absolutely without recourse.

Ever felt somewhat powerless, troubleshooting your internet connection with some underpaid cable company support rep? Now picture the same thing if you were an ant and the support rep a 100 foot-high concrete wall, and you may have an accurate allegory of dealing with Google or Facebook as an end-user.

Allow me to illustrate with two absolutely-true real-life examples of the hopeless situations one deals with, when a glitch occurs in the Matrix:

Continue reading

Freshly back from two weeks in Kyoto where I attended a Machine Learning Summer School organised at Kyoto University. The whole event was impeccably organised and featured some of the biggest names in the field lecturing on fundamental aspects of machine learning: two highly productive weeks leaving barely enough time in the evening to catch up with friends and enjoy life in my former hometown…

A few random thoughts and observations:

Continue reading

Can we instigate a rule wherein any internet company with more than 10 users that is found not to be using salted encrypted hashes for their user password database… gets to have its website shut down, servers sold for scrap and entire web development team slowly impaled on sharpened Aeron chairs?

People keep harping on the stupidity of end-users in their choice of passwords, but with proper hashing and salting, even password123 would make a halfway-decent password.

Rekindling with childhood activities fifteen-some year later… Turns out unlike bicycle, you do forget ice-skating. Couple hours later I was merely happy to stay up and move forward at a reasonable speed without breaking any bones. We’ll leave the figures for next year…

During Winter, Kyoto Aquarena (near Saiin, in the South-West of the city) turns its very large swimming pool into an ice-skating rink. Entrance is ¥1400 and skate rental is ¥600.

Pro-tip: show up after 5 on weekends and you get discount entrance and considerably less kids running into you at every turn (crowd thins out after 4 and they cut the ice again at 5).





If like me you deal with your typical Japanese administration office on a regular basis, you probably receive your fair share of documents, some of them occasionally packaged as a Zip archive

If also like me, you are not using a Windows machine, but a Mac running OS X or some flavour of Linux, you routinely end up with files bearing such poetic names as “Åuäwà ò_ï∂ä÷åWèëófiíÒèoìÕ.pdf”, “äwà ê\êøìÕÅyÉfÅ[É^Åz.xls” etc. This is due to some incompatibility between the way each system stores Japanese characters1To be specific: Windows seems to be using good-old antiquated Japanese-only SJIS, whereas OS X and others prefer spiffy universal encodings like UTF-8. and the fact the Zip format was never conceived to handle such differences. Not a big problem if you have one file, bit tedious if the archive contains 300 of them.

In the spirit of sharing the fruit of my last productivity-sink effort to fix that problem, I present you with a small script that takes such a Zip archive as input and correctly extract all the files (with their properly encoded filenames):

Continue reading

A very nifty trick I discovered while working on making KanjiBox accessible to blind users.

I previously mused that an iPhone/iPod made a much better and more cost-efficient language-studying tool than any dedicated electronic gizmo out there. This is now a thousand times truer…

One of the coolest features brought by version 4 of iOS (the software that runs on all iPhone/iPad/iPod Touch) is also one of the least known and used: VoiceOver is a built-in screen-reader geared at making Apple devices accessible to blind and visually impaired users. If you are such a user, you know about it already and will learn nothing here. For everyone else, this feature still has much to offer!

VoiceOver supports a dizzying collection of languages: from English to Japanese, Mandarin and Cantonese, most European languages, hell, even regional accents (English comes in US-, Brit- and Aussie-accented variants… Canadian-French as well). While the quality for English is about what you would expect from late-90s speech synthesis, the quality of some other languages is vastly superior. This is particularly true of the Japanese and French voices. To my very limited ear, Mandarin and Mexican-Spannish also sound quite close to human quality (Spain-Spanish, on the other hand, is pretty robotic).

As it turns out, your iPhone (/iPad/iPod Touch) comes with a native pronunciation teacher, out of the box. For hard-to-read languages like Japanese or Chinese, it can be a life-saver: helping you decipher SMS, emails or web pages, instead of relying on clunky, time-consuming, copy-pasting to a dictionary app.

Below are detailed instructions on how to enable VoiceOver and use it to read any text in any language on your iPhone (setup should be near-identical for iPads and iPod Touchs):

Continue reading