trevmex's tumblings

A great day at Disney!

Tue, 14 Nov 2023 19:50:41 -0500

Mickey and Minnie visited the Seattle Disney Office for the first time in 18 years, and though Disney VoluntEARS and US Hunger we fed 40,000 people through Northwest Harvest!!!

Years ago, when I was just a manager, I was (and still am) passionate about tech meetups and sharing…

Tue, 05 Sep 2023 13:33:30 -0400

Years ago, when I was just a manager, I was (and still am) passionate about tech meetups and sharing information. So, I organized a tech meetup at our offices (I was at a large media company at the time) for after hours. This is something I normally did. Many months of me doing this went by, and finally the office admin lashed out at me in frustration. They said, “don’t you know that when you do things off the cuff, like plan an after hours meeting, you are creating extra work for so many people?!? You should really think about how your actions affect others before leaping in.”

I still have a bad habit of leaping in, though I have reflected often on that experience and it has shaped my world view on the ripple effects my actions cause in the world.

Self-deprecation: Story Time

Fri, 19 May 2023 13:24:06 -0400

Let me tell you a story.

I was at a tech meetup once and raised my hand to ask a question. I said, “This is a dumb question, but …” and asked my question.

I didn’t think anything of it. After the meetup, a good and always straightforward friend of mine came up to me and said, “Don’t call your questions dumb.”

To which I said, “Oh, but it was just my ignorance…”

He stopped me. He said, “No, you don’t understand. In that room, people look up to you. I know some people didn’t even know to ask that question. And others that had the same or similar questions. When you deprecate yourself around others, you de facto deprecate those around you thinking the same thing. Don’t self-deprecate in front of others. It adds nothing and lowers the self-esteem of those around you.”

I was blown away. I never thought of self-depreciation as harming others, but he is right. Knowledge, skill, talent: it is a spectrum. And when I self deprecated in front of others, that negativity has impacts beyond just me.

Let’s say you self-deprecate about what you do. Are you the best in the world? Probably not. You know that. Are you better than a hell of a lot of people? Yes. Yes, you are. Who does that self-deprecation serve? Not me.

Pistachio Brittle...

Fri, 23 Dec 2022 17:29:10 -0500

Pistachio Brittle making!
https://www.instagram.com/p/Cmhvk1Dv8cZ/?igshid=NGJjMDIxMWI=

Great lean coffee topics for the last two mornings of Strange...

Sat, 24 Sep 2022 12:05:51 -0400

Great lean coffee topics for the last two mornings of Strange Loop!!! (at St. Louis, Missouri)
https://www.instagram.com/p/Ci5UKakOPsd/?igshid=NGJjMDIxMWI=

LEVERAGING AI FOR GLOBAL HEALTH: BENSHI.AI’S DATA-CENTRIC MACHINE LEARNING PLATFORM

Sat, 24 Sep 2022 11:18:32 -0400

Africa Periañez

BENSHI.AI, CEO

Leveraging AI for global health: benshi.ai’s data-centric machine learning platform - Strange Loop

Focusing on bringing AI to the most underserved countries in the world.

40% countries have fewer than 10 medical doctors per 10,000 people!

HIV/AIDS is the leading cause of death across the world, especially in Africa.

There are many cheap, easy-to-implement technologies that can improve health outcomes across the world.

Most of the problem isn’t getting the supply out there, it is getting people to understand, and trust, that the medicine is safe and trustworthy.

The name of the game, then, is engagement. We need to engage folx that do not know what is safe or not.

A great way do this is through games!

Benshi’s were people that would describe silent movies to illiterate people in Japan. Origin of the name benshi.ai

They want to democratize medical knowledge through mobile phones. Mobile phone penetration is widespread. 80% of people in Africa have smart phones connected to the Internet.

Global health is a messy field that is not remotely organized globally, so there is a lot of communication needed at the local level to bring people together.

They focus on experimentation and machine learning to forecast health trends

The goal is to predict and nudge behavior towards improved health decisions.

They work to gamify and personalize interventions and predictions for each client.

It is not enough to have data, you need to have organized data to mine the value out of it.

Benshi uses data labeling to to organize data and build data sets, which feed back to client nudges

Benshi is n opinionated system. They want to guide clients towards decisions that will save their lives.

A Nigerian doctor can visit about 100 patients a day. In Ethiopia, there is 1 midwife per 10,500 people.

How can we use AI to optimize the limited resources of medical professionals to triage the highest need cases in within these constraints?

Malaria is 100% preventable, treatable, and curable, but we still see 200,000 deaths every year. We want to nudge people to take their malaria medication through gamification.

If a person goes to a pharmacy, and they do not have the drugs they need right away, there is a 29.4% chance that person will never return to get their medicine in developing countries.

Part of what we can do is predict the medicine needs per area using AI modelling.

ML helps us discover and predict actual demand and needs per area. That can kickstart the supply chain ahead of the resource needs

Big Pharma is not incentivized to participate in these improvements right now. So another problem is how do we encourage and incentivize Big Pharma to lean in as well?

There are very few people working in this field, despite this being a very large problem for the world

M. Kramer, E. Duflo, and A. Banejee got the Nobel prize for work in this field, and much more is needed in experimentation in low income countries and global health

Check out Milind Tambe’s Field Study in Deploying Restless Multi-Armed Bandit in India

Tracking out of stock medicines over the world help s drive preemptive resource management

The more people that feed data into ML model apps, the better refines the predictions have gotten in Indonesia.

The team is largely woman-driven, 83% in leadership, 55% in engineering

Thank you, Africa, for the great talk!

Strange Loop about to start (at St. Louis,...

Fri, 23 Sep 2022 19:07:18 -0400

Strange Loop about to start (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3fmm-JJKL/?igshid=NGJjMDIxMWI=

Gjeta Gjyshinca making monads easier at Strange Loop (at St....

Fri, 23 Sep 2022 19:05:14 -0400

Gjeta Gjyshinca making monads easier at Strange Loop (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3fXdepWhZ/?igshid=NGJjMDIxMWI=

John Romero speaking at Strange Loop (at St. Louis,...

Fri, 23 Sep 2022 19:03:50 -0400

John Romero speaking at Strange Loop (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3fNKSpuuc/?igshid=NGJjMDIxMWI=

Bruce Eckel talking polymorphism at Strange Loop (at St. Louis,...

Fri, 23 Sep 2022 19:01:51 -0400

Bruce Eckel talking polymorphism at Strange Loop (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3e-e5J9s4/?igshid=NGJjMDIxMWI=

Philip Haller talking about Spore3 at Strange Loop (at St....

Fri, 23 Sep 2022 19:00:28 -0400

Philip Haller talking about Spore3 at Strange Loop (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3e0duJ0wY/?igshid=NGJjMDIxMWI=

julia ferraioli and amanda casari talking about how OSS has...

Fri, 23 Sep 2022 18:58:49 -0400

julia ferraioli and amanda casari talking about how OSS has broken the internet multiple times. (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3eoigp-7s/?igshid=NGJjMDIxMWI=

The vendor scene at Strange Loop! (at St. Louis,...

Fri, 23 Sep 2022 18:55:11 -0400

The vendor scene at Strange Loop! (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3eN6npljT/?igshid=NGJjMDIxMWI=

@horrorcheck presenting on Smooch, the modern KiSS doll...

Fri, 23 Sep 2022 18:53:38 -0400

@horrorcheck presenting on Smooch, the modern KiSS doll system!!! (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3eCeIJKaJ/?igshid=NGJjMDIxMWI=

Alex Miller, the founder of the Strange Loop conference. (at St....

Fri, 23 Sep 2022 18:50:59 -0400

Alex Miller, the founder of the Strange Loop conference. (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3dvDNpwRP/?igshid=NGJjMDIxMWI=

@exploitsinbaking taking a break at PWLConf. (at St. Louis,...

Fri, 23 Sep 2022 18:35:41 -0400

@exploitsinbaking taking a break at PWLConf. (at St. Louis, Missouri)
https://www.instagram.com/p/Ci3b-1uvHV6/?igshid=NGJjMDIxMWI=

ID – THE EARLY DAYS OF ID SOFTWARE: PROGRAMMING PRINCIPLES

Fri, 23 Sep 2022 18:17:51 -0400

John Romero

ROMERO GAMES

Id – The Early Days of Id Software: Programming Principles - Strange Loop

John started learning BASIC as an elementary school kid. Then got an Apple ][. He eventually learned 6502 Assembly.

Made software for the Air Force in High School, since his dad worked in the Air Force.

At 21, John was working porting games to the Commadore 64.

Made a tool to make games. Hired John Carmack, Adrian Carmack and Tom Hall to make games.

Carmack ad Hall made Dangerous Dave in Copyright Infringement, Sept. 20, 1990. The demo proved you can do smooth scrolling on a PC!!!!!

Commander Keen was the first PC game that could do smooth scrolling horizontally. The Commander Keen trilogy was made in three months. So popular people cosplay as commander keen.

Started licensing the Keen engine in 1994.

ID Software was started in Shreveport, Louisiana.

They played D&D on the weekends, Quake was inspired by one of there D&D sessions.

They made a SMB3 PC demo to pitch to Nintendo, but Nintendo (wisely) decided to only allow their software to run on their hardware.

They always were making two games at a time. No prototypes, always maintain contently shippable code. Only make games, not prototypes of games. It was a small team of four people.

They were making games 24/7, and they moved from Louisiana to Wisconsin.

Your game has to be able to be run by your team at all times. Bulletproof your engine by providing defaults upon load failure.

If you fail to load a sprite, load a bagel. If the sound fails, play an annoying sound.

Keep your code simple. Always make it simpler all the time.

After six months of snow, they went to Texas.

Great tools help make great games.

Wolfenstein 3D took 4 months for the first episode. 6 months for the rest.

Licensed the Wolfenstein engine for Shadowcaster.

We are our own best testing team. There is no throwing it over the fence. There is no QA, just you!

As soon as you see a bug, you fix it! If you don’t fix your bug, new code will be built on buggy code.

Doom was based on their D&D campaign, Evil Dead, and Alien.

Use a development system that is superior to your game.

Doom was developed in NextStep systems.

Doom was a 5 person team. Then Tom Hall left, and they hired two more people.

In the middle of building Doom, they had to port Wolfenstein 3D for the Super Nintendo. It took them three weeks.

Doom was INSANELY popular.

The final day of Doom’s creation the team worked 30 hours straight. On some computers, the game just froze. Carmack thought about the bug, and fixed it.

Doom II took them 8 months!

Then John made Heretic.

1995, they started working on Quake. With 9 developers/designers.

No code in Doom was used in Quake. Write this code for this game only, not for future games. You are going to be a better coder in the future.

They planned to make Quake on a Cray 64 super computer. Unfortunately Cray was bought by Silicon Graphics and the agreement fell through.

John then made Hexen, but Hexen II was not the same system.

Strife was the very first FPS RPG.

In 1995, 9 developers, released 2 new games.

Created Qtest around this time.

Encapsulate functionality to ensure design consistency. This minimizes mistakes and saves times.

June 22, 1995, 5 pm CST Quake shareware was released in university FTP.

The Pentium FDIV bug was discovered with Quake.

Quake was the first internet multiplayer, Quakeworld was launched, first mouse driven FPS.

Try to code transparently. Coders are not black boxes. Coding is a creative art form based in logic.

They didn’t have source control!!!

Thank you, John, for the great talk!

REVIVING 1990S DIGITAL DRESS-UP DOLLS WITH SMOOCH

Fri, 23 Sep 2022 17:27:04 -0400

Libby Horacek

POSITION DEVELOPMENT

@horrorcheck

Reviving 1990s Digital Dress-Up Dolls with Smooch - Strange Loop

What is the Kisekae Set System: It is a system to make digital dress up dolls!

Created in 1991 by MIO.H for use in a Japanese BBS. Pre-web!!! By separating the system from the assets of the dolls, you can make the systems much smaller

CEL image format. They have a transparent background and indexed colors like GIFs!

There is also the KCF Palette Format. Each GUF stores its own palette, with CELs there is a shared KCF palette file. Which is a file size cost savings. You do not have to repeat your color info per file.

Having your palette in another file makes palette swapping really easy! Just swap the KCF file!

CNF configuration files dictate layering, grouping, setting, and positions!

KiSS dolls have a lot of files, so they used LhA, the most popular compression format in Japan at the time.

In 1994, KLS (a user) created the KiSS General Specification

1995 FKiSS is born by Dov Sherman and Yav, which adds sounds, and animation!

FKiSS 2, 3, 4, two versions of FKiSS 5! So much innovation!

1995-2005 huge growth in KiSS! Increase of access, mainstreaming of anime, younger and more female audience.

2007-2012: KiSS declined due to it being much harder to make. All the old tutorials were written for older systems, English-speaking KiSS-making died out by 2010

The Sims, and other doll making was more accessible.

But, why KiSS? If I can make dolls in other places. They are great snapshots of the pre-internet world, and how play online evolved early on.

Lots of fun dolls were available, and it would be cool to save it.

Picrew is a modern thing people use to make dolls.

Tagi Academy is a tamagochi game within KiSS. Impressive!

KiSS has an open specification!!!! That is super cool! That means anyone can make your own viewer, as opposed to a closed system in The Sims.

So, why NOT make a KiSS interpreter?!!

Libby made Smooch, a KiSS renderer written in Haskell, which at the Recurse Center!

Smooch used a web framework called Fn (fnhaskell.com)

Had to make a CNF file parser using Parsec library that uses parsing combinators.

She created a data type that houses everything that can be in a CNF file, and parses it correctly in priority order.

Parsing is a great candidate for test-driven development. You can write a test with a bad CNF file, and then make sure your parser handles it.

The parser translates CNF lines into JSON. Uses the ASON library to translate into JSON.

First tried cel2pnm, coded with help from Mark Dominus at Recurse Center Made a C program that converted cels to portable bitmaps, which could be translated into PNGs.

Then JuicyPixels was created to translate palette files directly in Haskell

Now it is converted to JavaScript! No libraries, just JS!

Using PNGs in JS, thought, made it hard to click on parts of the clothing! Since it is squares.

So, you use ghost canvases! You use tow canvases, one on top of the other, to find the color of the pixel you clicked on, If you clicked on a scarf color, it will pick the scarf, and if you pick the sweater, you get the sweater!

Libby just added FKiSS 1 to Smooch! So we have animations now! The animations are basically event-based scripting the CEL files.

Smooch translates FKiSS to JSON, then the interpreter translates the JSON to JavaScript!

An action in FKiSS is translated into a function in JS. To do this you have to use bind in JS.

The events become CustomEvents in JS! So it looks like a regular event on the DOM.

What is the future of KiSS? Let’s get more people making KiSS dolls!!! So why not make it easy to make dolls using PNGs.

How can we make people interested in building KiSS dolls and KiSS tooling.

Smooch need contributors!

GitHub - emhoracek/smooch: Kisekae Set System implementation for the web

Smooch!

Thank you, Libby, for the great talk!

REMEMBER WHEN WE BROKE THE INTERNET?

Fri, 23 Sep 2022 16:09:14 -0400

REMEMBER WHEN WE BROKE THE INTERNET?

julia ferraioli

OPEN SOURCE TECHNICAL LEADER @ CISCO

amanda casari

OPEN SOURCE SCIENTIST @ GOOGLE

@juliaferraioli, @amcasari

Remember when we broke the internet? - Strange Loop

There are times in our past when open source, or something like open source broke the internet. It is ALL interconnected!

In Open Source, you cannot restrict who uses your technology.

Complexity Theory: unpredictable systems that operate by independent rules. And the emergent system is more complex that its parts.

A sociotechnical system is a complex system that contains people as nodes in the system.

Open Source is a complex sociotechnical system. OSS is fundamentally a human-based system. We have to take in consideration how we as humans interact with each other

A “black swan” event is an evet that disrupts the status quo, causes systematic change, and seem inevitable in hindsight.

Context matters. You have to understand the history behind where systems have evolved over time. And where evolution happens as a disruption of norms.

The Morris Worm 1988 - Created by Robert Tappan Morris at Cornell to attack MIT by exploiting transitive trust between sys admins and users.

Morris was the first arrest under the computer fraud act of 1986

We do not assume, now, that everyone is operating in everyone else’s best interests. Morris himself helped spread the solution to the worm.

Super homogeneity is now though as a risk.

Heartbleed 2014 - Heartbleed was released is 2012, but publicly announced in 2014. Security flaw in OpenSSL (used EVERYWHERE is TLS). 66% of the internet used OpenSSL.

Heartbleed increased funding towards developers and maintainers of open source software.

Corporations start paying people to work on open source software.

left-pad 2016 - trademark disagreement over an npm package name. The maintainer deleted ALL his libraries, including left-pad, and broke a TON of services.

Npm published the left-pad package without permission of the main maintainer. It is “free to Fork” which should be OK.

Choose your license carefully.

This increased visibility of the cost on maintainer burnout.

New issues abound on maintainer rights.

Fauxpen 2018 - Licensing wars… Companies start using open source as a business strategy. OSS used to be counter culture, now it is corporate culture.

Companies developed OSS n order to monetize it. And at the same time the cloud was born.

The companies that made the OSS started reversing their licenses. Example: Akka.

A haiku:

A need for income

The actors have to get paid

My old friend, Akka

Open source waters are muddied: Open Core is being a thing, where the OSS is a core that is useless, but still open.

Exposed OSS as a business strategy. How can companies be a part of the ecosystem ethically.

Also, capitalism introduced into OSS is a sticky mess.

#TODO(everyone) https://todogroup.org/

OSS is a sociotechnical system. OSS is lost to walled gardens, private conversations, and an overindex on source code. OSS needs to be more transparent.

Glue work is critical, but not recognized. Feature work is given more visibility than glue work for systems, though it is the critical and essential work of OSS.

Our OSS narratives are not complete, they are overindexed by a cult of personality. As long as OSS is captured by hero engineers and cult of personality founders, the system will be broken and glue work will be undervalued.

This is a sustainability problem. We didn’t light this fire, but we are here to fight it!

Thank you Julia and Amanda!

HOW TO AVOID SAFETY HAZARDS WHEN USING CLOSURES IN SCALA

Fri, 23 Sep 2022 15:22:21 -0400

Philipp Haller

ASSOCIATE PROFESSOR AT KTH ROYAL INSTITUTE OF TECHNOLOGY, SWEDEN

@philippkhaller

How to avoid safety hazards when using closures in Scala - Strange Loop

Closures are essential in Scala. Map and reduce are example of functions that are based off of closures.

Closures are also important for concurrent programming in Java.

If you are using Futures in Scala, you are also using Closures.

BUT, it is also easy to get things wrong while using closures.

A distributed data set is set across multiple data nodes, and the closure has to be serialized across all nodes.

Safety issues stem from unrestricted variable capture within closures. So we could deeply clone an object before using it in the closure as a potential solution.

How do you spot unsafe closures?

Capturing vars (reassignable variables)
In Java, captured variables need to be final, but not in Scala
Mutable types in closures can be an issue
Types that are not serializable

How to write safer closures: verify the creation of the closure

What is the logical snapshot of the memory that the closure should be initialized in?

Verify the semantics of the closure’s execution. Does it mutate? Is it transactional? (hint, it shouldn’t be)

Introducing Spores3, a new Scala 3 library to prevent these pitfalls.

Spores3 provides a dropin abstraction for closures

Spores3 is a complete reimplementation of spores from Scala.

Macro usage is simple and robust. Designed to be portable from the beginning, as opposed to spores for Scala 2.

val s = Spore((x: Int) => x + 2)

The type is Spore[Int, Int] { type Env = Nothing }

Nothing is the bottom type in Scala. No type has a type of Nothing.

Spore types are subtypes:

sealed trait Spore[-T, +R] extends )t => R) { type Env }

The Env type member of the Spore trait expresses type-based constraints.

A spore parameter can only capture thread-safe types.

We need a flexible way of serializing closures to make them portable. Serialization needs to be determined statically for safety.

Serialization is used for communication, not storage.

Every node needs to be running the same code

We also don’t want to pass byte code back and forth.

In practice, we create a names Spore.Builder to create the spore to make a SporeData to “pickle” the spore (aka serialize it).

SporeData factory uses a macro to check for safety!

GitHub - phaller/spores3: Abstractions for making closures in Scala safer and more flexible

Thank you, Philipp, for the great talk!