Twitter Archive: 2021
Twitter dot com was a microblogging and social networking service, where users could share short posts (commonly known as “tweets”) with text, images, and video.
This is an archive of all my tweets from 2008–2025.
These days, you can find me on Bluesky or Mastodon.
By year
- 2025 33 posts
- 2024 203
- 2023 97
- 2022 91
- 2021 89
- 2020 26
- 2019 0
- 2018 6
- 2017 56
- 2016 108
- 2015 71
- 2014 197
- 2013 180
- 2012 160
- 2011 162
- 2010 171
- 2009 238
- 2008 101
…or see popular posts.
2021
Ways of configuring a CLI tool:
1️⃣ Command-line options
2️⃣ Config file
3️⃣ Env vars
4️⃣ Comments in input files (e.g. eslint, pretter)
Huge difference. In 3 & 4 you don't need to know hardly anything about how the process is spawned, etc. It "just works"
1️⃣ Command-line options
2️⃣ Config file
3️⃣ Env vars
4️⃣ Comments in input files (e.g. eslint, pretter)
Huge difference. In 3 & 4 you don't need to know hardly anything about how the process is spawned, etc. It "just works"
I think this is a big part of why it's so much easier to use console.log vs. a debugger. Debugger often requires you to modify the way the process is launched. Adding a console.log / printf just works, even if your process is remote (e.g. CI), spawned by the test framework, etc.
Using "just JS" (not TypeScript) for side projects makes me feel so ~free~
…until I waste 30 minutes debugging something that would have been a type error.
…until I waste 30 minutes debugging something that would have been a type error.
This is just as much a condemnation of the state of debugging tools as it is a pro-types thing.
If it had taken only 1-2 minutes to debug, the mistake wouldn't have bothered me at all.
If it had taken only 1-2 minutes to debug, the mistake wouldn't have bothered me at all.
🤩 I love this! See @MockMechanics for more examples. twitter.com/MockMechanics/…
Imagine if you could create debug visualizations like this as easily as using `console.log` twitter.com/steveruizok/st…
Feels good to see a paper that is a direct response to (and improvement on) something I published. Makes me feel like I actually put a brick in the wall of knowledge 😊 twitter.com/_ohmjs/status/…
What are the best ways of handling versioning and experimental features in a programming lang?
• Python has the __future__ module: python.org/dev/peps/pep-0…
• JS has 'use strict'
• Handle it outside of the language. Rust (?), JS (script vs. module)
• Python has the __future__ module: python.org/dev/peps/pep-0…
• JS has 'use strict'
• Handle it outside of the language. Rust (?), JS (script vs. module)
Dart's approach seems really well thought out: github.com/dart-lang/sdk/… (thanks @mraleph)
One thing I am curious about is why experiments can't be enabled by the @ dart comments.
E.g. `// @ dart=2.9+non-nullable` doesn't seem to be valid, only `// @ dart=2.9`.
E.g. `// @ dart=2.9+non-nullable` doesn't seem to be valid, only `// @ dart=2.9`.
Seems like a missed opportunity that you can't click and drag the blocks in the Scratch documentation just start creating your program there.
This makes me reconsider the advantages of structure editors (e.g. hazel.org) and blocks-based programming environments like Scratch.
It's usually considered a benefit that you can't make certain errors, but maybe there is also a lost opportunity there? twitter.com/evincarofautum…
It's usually considered a benefit that you can't make certain errors, but maybe there is also a lost opportunity there? twitter.com/evincarofautum…
It's similar to the problems with disabled buttons / greyed out menu items. As you user, you can see that something is not allowed, but you can't figure out why! smashingmagazine.com/2021/08/frustr…
It's interesting to think about languages / environments that would actually encourage you to break the rules, in order to teach you what the rules are.
Oooh —
"Error training proved to lead to higher performance than a training that reduced the chances to make errors, the so-called error-avoidant training"
core.ac.uk/download/pdf/2…
"Error training proved to lead to higher performance than a training that reduced the chances to make errors, the so-called error-avoidant training"
core.ac.uk/download/pdf/2…
A fantastic overview of parsing techniques and the practical challenges & benefits of each. twitter.com/laurencetratt/…
"Dynamic, in-air device formations"
Impressive work from @nicmarquardt et al! Back in the @bumptop days we thought about doing something like this, but it wasn't really feasible with the iPhone 3G.
youtube.com/watch?v=WVEEsi…
Impressive work from @nicmarquardt et al! Back in the @bumptop days we thought about doing something like this, but it wasn't really feasible with the iPhone 3G.
youtube.com/watch?v=WVEEsi…
This has been a lot of work, excited to finally release it! twitter.com/_ohmjs/status/…
listening to *extremely 90s music* and working on the @_ohmjs v16 release. how's your friday
Great thread. Related to the notion of "winner's games" and "loser's games": junto.investments/losers-game/ twitter.com/danluu/status/…
Why can’t I use emoji reactions on paragraphs in Google Docs yet
Is there a name for this pattern popularized by React?
1️⃣ Produce a description of the desired state of a mutable resource, and
2️⃣ a reconciler which mutates the resource to match the description.
Terraform is also similar.
IMO it's also a nice pattern for making testable APIs
1️⃣ Produce a description of the desired state of a mutable resource, and
2️⃣ a reconciler which mutates the resource to match the description.
Terraform is also similar.
IMO it's also a nice pattern for making testable APIs
The Command pattern is similar in some ways, and can achieve some of the same goals, but with CP you:
1️⃣ Produce a list of actions (commands) to achieve the desired state, and
2️⃣ interpret the commands and perform the actual mutation.
1️⃣ Produce a list of actions (commands) to achieve the desired state, and
2️⃣ interpret the commands and perform the actual mutation.
Ok, I guess it's called the *reconciler pattern*: oreilly.com/library/view/c…
What are some other examples of this pattern? I'm curious whether it originated in React, or somewhere else.
What are some other examples of this pattern? I'm curious whether it originated in React, or somewhere else.
Reminds me of the great paper OOPSLA '04 paper "A Unified Theory of Garbage Collection": web.eecs.umich.edu/~weimerw/2012-… twitter.com/pcwalton/statu…
I ❤️ these "design tool micro-UX" threads from @steveruizok. twitter.com/steveruizok/st…
Not sure if this is descriptive or prescriptive. twitter.com/mraleph/status…
Interesting! One of those results where you think, "Ohhh, that makes a lot of sense" as soon as you hear it. twitter.com/AdamMGrant/sta…
satisfying
What new UI patterns for social/collaborative software have appeared in the past 5-10 years, that will stick around for another 20 years?
My bets:
• Emoji reactions (Slack, iMessage)
• Comment threads + "Resolve conversation" (GDocs, GitHub)
Others?
My bets:
• Emoji reactions (Slack, iMessage)
• Comment threads + "Resolve conversation" (GDocs, GitHub)
Others?
Another app jumping on the "Resolve conversation" train 😁
twitter.com/typefully/stat…
twitter.com/typefully/stat…
Somewhat surprising that TypeScript doesn't flag this as an error. (And yes, `strictFunctionChecks` is enabled.)
I wonder why?
Playground link: typescriptlang.org/play?noImplici…
I wonder why?
Playground link: typescriptlang.org/play?noImplici…
I'm guessing it's because the type of `this` can only be observed from inside the function. So it's not unsafe to allow the use of an arrow function here, and it would overly restrictive to disallow it.
The first parser I ever wrote was in Squeak. It also sparked my interest in VM implementation, which eventually led to me working on the J9 Java VM (en.m.wikipedia.org/wiki/OpenJ9) for several years, which was an incredible learning experience. twitter.com/SqueakSmalltal…
Cyberspace isn't cool…you know what's cool?
« Le cyberespace »
« Le cyberespace »
Interesting 🧵, some papers that are now on my to-read list. twitter.com/sliminality/st…
Yesterday, I was using Google Maps in the car and it announced an alternate route would save “six megawatts” (screen said 6 mins). 🤔
Anyone else experienced a text-to-speech glitch like this? I’m super curious to know how this could happen.
Anyone else experienced a text-to-speech glitch like this? I’m super curious to know how this could happen.
Another time it told me that my destination was on the right, but pronounced it “destinación”. So weird.
This is why I fed conflicted about the idea of dynamic media replacing papers, textbooks, etc.
I routinely read papers that are 30 or 40 years old. When 30-40 year old software is easily runnable, it’s a wonderful surprise. twitter.com/spiralganglion…
I routinely read papers that are 30 or 40 years old. When 30-40 year old software is easily runnable, it’s a wonderful surprise. twitter.com/spiralganglion…
Relevant 👉 The Cuneiform Tablets of 2015: vpri.org/pdf/tr2015004_…
“We discuss the problem of running today’s software decades, centuries, or even millennia into the future.”
“We discuss the problem of running today’s software decades, centuries, or even millennia into the future.”
I meant to call out the great work that @internetarchive is doing here. twitter.com/utopiah/status…
Also, @codefrau’s SqueakJS has been super helpful to me: squeak.js.org
Interesting 🧵 about the finer points of “liveness” in programming environments, and how they affect the PX. twitter.com/lukego/status/…
Love this metaphor. twitter.com/devonzuegel/st…
Falsehoods programmers believe about age.
Bert Sutherland's thesis, "The On-line Graphical Specification of Computer Procedures" from 1966: dspace.mit.edu/bitstream/hand…
As far as I know, this is the oldest example of a nodes-and-wires, graphical programming language. Is there something earlier?
As far as I know, this is the oldest example of a nodes-and-wires, graphical programming language. Is there something earlier?
Gonna start putting this on page 1 of all my papers.
Been thinking…there are lots of great resources these days that teach you how to create compilers and interpreters.
But — what are the best resources for creating a visual programming environment?
But — what are the best resources for creating a visual programming environment?
Same with live programming. I feel that there is a bunch of accrued knowledge in the heads of folks who have been doing this for years, but very little (that I know of) that’s documented and accessible to newcomers.
This is superb advice. twitter.com/b0rk/status/14…
It's funny how we build up these tiny situated vocabularies to refer to things in our house, neighbourhood, etc.
In a new house, there's this adjustment period — do we call this the "office", "spare room", "guest room"?
In a new house, there's this adjustment period — do we call this the "office", "spare room", "guest room"?
After a while you forget that it's not really legible to others. Then you have someone staying with you, and you realize that the words you use every day are ambiguous or even illegible to them.
In my family it's even more complex, because we use the German word for some rooms, items, etc. and the translation to English isn't even obvious. Like "Bauernschrank" for instance.
"The language Lisp has changed repeatedly, each time becoming a new genre." twitter.com/fogus/status/1…
This looks super useful! I remember when I first started using Linux, I wanted something like this for writing shell pipelines. twitter.com/rsnous/status/…
In an age of ubiquitous high-res displays, this makes a whole lot of sense. twitter.com/rsms/status/14…
Whoah. twitter.com/mattparrilla/s…
I had the chance to help develop Sourcegraph's career levels for engineering ICs. It was super challenging, but also a lot of fun! twitter.com/nickdsnyder/st…
Love this…and surprised I haven't seen it before. Paper from UIST '09: dspace.mit.edu/bitstream/hand… twitter.com/alltom/status/…
A little project for my parental leave this summer. twitter.com/_ohmjs/status/…
Last day of my parental leave — this project is coming in under the wire! 😆
twitter.com/_ohmjs/status/…
twitter.com/_ohmjs/status/…
I'm biased, but @Sourcegraph is so useful for open source maintainers.
I want to deprecate a method in Ohm, and I'm wondering how many people use it. A quick Sourcegraph tells me: not many.
sourcegraph.com/search?q=conte…
I want to deprecate a method in Ohm, and I'm wondering how many people use it. A quick Sourcegraph tells me: not many.
sourcegraph.com/search?q=conte…
I wish GitHub offered a more collaborative way of working with pull requests from new contributors.
Allowing the maintainer to push commits to the PR branch enables a more "yes, and…" approach, but GH doesn't make this easy.
Allowing the maintainer to push commits to the PR branch enables a more "yes, and…" approach, but GH doesn't make this easy.
For a recent Ohm PR from @joshmarinacci, I tried this:
1️⃣ Create a new branch in the dest repo
2️⃣ Merge the original PR into the new branch
3️⃣ Open a new PR from the new branch
Now both the maintainer and contributor can work on the PR. Is there an easier way?
1️⃣ Create a new branch in the dest repo
2️⃣ Merge the original PR into the new branch
3️⃣ Open a new PR from the new branch
Now both the maintainer and contributor can work on the PR. Is there an easier way?
Woody Guthrie vibes
Not sure why it took me so long, but I finally created a Twitter account for Ohm: twitter.com/_ohmjs.
Hit the follow button to keep up with new features, releases, etc.
Hit the follow button to keep up with new features, releases, etc.
Great thread about the development of visual language, which is something I've been fascinated to watch in my (now) 5 year old. t.co/qQEUYYHopF
This has been my team’s main focus for the past 8 weeks or so…psyched to see it ship! Plus the new design still makes me smile every time I see it. twitter.com/Sourcegraph/st…
Also discussed — a nice example of an arithmetic-to-WASM compiler using Ohm: guitarvydas.github.io/2021/05/15/WAS… twitter.com/warianoguerra/…
Every subject should have a series like this.
Huh, guess I wasn't the only one. twitter.com/rineshthomas/s…
This aesthetic
*rubs temples*
Thinking a lot about this.
When I first saw Google+, I realized that the concept of "circles" totally missed the fact that Twitter, Facebook, etc. are different *places*, not just access control lists. twitter.com/anselm_io/stat…
When I first saw Google+, I realized that the concept of "circles" totally missed the fact that Twitter, Facebook, etc. are different *places*, not just access control lists. twitter.com/anselm_io/stat…
I mean the real wtf here is that JSON doesn’t support comments twitter.com/aboodman/statu…
Is there a name for the abstract concept behind…
Jevon's paradox: en.wikipedia.org/wiki/Jevons_pa…
risk compensation: en.wikipedia.org/wiki/Risk_comp…
and induced demand: en.wikipedia.org/wiki/Induced_d…
?
They seem like instances of a more general phenomenon.
Jevon's paradox: en.wikipedia.org/wiki/Jevons_pa…
risk compensation: en.wikipedia.org/wiki/Risk_comp…
and induced demand: en.wikipedia.org/wiki/Induced_d…
?
They seem like instances of a more general phenomenon.
Like, we observe a causal relationship x → y. So we change x, but achieve only a temporary effect on y, due to latent variables that we didn't consider / weren't aware of.
Seems like there must be a name for it in system dynamics / systems theory.
Seems like there must be a name for it in system dynamics / systems theory.
A few people have mentioned Le Chatelier's principle.
twitter.com/shapeoperator/…
twitter.com/shapeoperator/…
Should languages require an explicit entry point (e.g. "main"), or not?
(I surprised myself by not having an opinion on this.)
(I surprised myself by not having an opinion on this.)
Most “scripting” languages don’t, and most compiled languages do, but is there a good reason for that?
It occurred to me that maybe the distinction is between langs that allow top-level expressions vs those that don't. E.g., Haskell → no top-level expr, has main().
The Swift doc even calls this out explicitly: developer.apple.com/swift/blog/
The Swift doc even calls this out explicitly: developer.apple.com/swift/blog/
See also twitter.com/vb6rox/status/…
My (long) weekend project: a Smalltalk-to-JS compiler using Ohm. 👉 github.com/pdubroy/ohm-som
Some progress this week{end} — it can load classes from the SOM stdlib (e.g. Integer), and evaluate basic expressions w/ primitive and non-primitive methods.
More progress — I can successfully execute at least some parts of the SOM test suite 🙌. Now I'm in the fun cycle of:
1️⃣ Run test suite.
2️⃣ It breaks.
3️⃣ Fix breakage.
4️⃣ GOTO 1
1️⃣ Run test suite.
2️⃣ It breaks.
3️⃣ Fix breakage.
4️⃣ GOTO 1
Tons of progress on my Smalltalk-to-JS compiler in the last few weeks.
It now passes the majority of the SOM test suite! 🎉 Just need to solve a few special cases now.
It now passes the majority of the SOM test suite! 🎉 Just need to solve a few special cases now.
Final achievement for the long weekend:
✅ DoesNotUnderstandTest
✅ DoesNotUnderstandTest
So Ohm was on the front page of HN yesterday: news.ycombinator.com/item?id=266033…. Mostly positive but of course the biggest thread was "why would you ever write a compiler in JavaScript" 🤦
Super thoughtful and interesting thread about code reviews. I’ve worried about many of these issues as well, especially after seeing the negative parts of async code reviews while I was at Google. twitter.com/searls/status/…
Two things that have strongly influenced how I think about code review:
- Richard Gabriel on writers' workshops: dreamsongs.com/Files/WritersW…
- Kurt Ralske's "The Crit" retnull.com/index.php?/tex…:
- Richard Gabriel on writers' workshops: dreamsongs.com/Files/WritersW…
- Kurt Ralske's "The Crit" retnull.com/index.php?/tex…:
Figma quietly advancing the state of end-user programming. 👀 twitter.com/rogie/status/1…
Some personal news:
This week I joined @srcgraph as an Engineering Manager on the Frontend Platform team.✨
I used Sourcegraph a *ton* in my day-to-day work at Lyft, and now I'm excited to help make it better for everyone!
This week I joined @srcgraph as an Engineering Manager on the Frontend Platform team.✨
I used Sourcegraph a *ton* in my day-to-day work at Lyft, and now I'm excited to help make it better for everyone!
This is an interesting perspective. Many things that look like hype cycles might actually be “effectiveness cycles”.
And we see this in lots of areas: programming languages, system architecture, etc. twitter.com/CasualEffects/…
And we see this in lots of areas: programming languages, system architecture, etc. twitter.com/CasualEffects/…
In the context of graphics, this reminds me of the theory of constraints: en.wikipedia.org/wiki/Theory_of…
I love all these different takes on online copresence that we're seeing.
I'm imagining a world where everyone you meet with has their own personal, idiosyncratic virtual environment you go to when you want to see them… twitter.com/waxpancake/sta…
I'm imagining a world where everyone you meet with has their own personal, idiosyncratic virtual environment you go to when you want to see them… twitter.com/waxpancake/sta…
Great thread to see two very different ways of looking at UX.
For me “Emotional Design” by @jnd1er was more eye-opening than The Design of Everyday Things. twitter.com/kocienda/statu…
For me “Emotional Design” by @jnd1er was more eye-opening than The Design of Everyday Things. twitter.com/kocienda/statu…
Very cool — “edit wear” for VS Code! twitter.com/Wattenberger/s…
This is awesome: “hills look steeper…when wearing a heavy backpack”.
Action-specific perception: en.m.wikipedia.org/wiki/Action-sp…
Action-specific perception: en.m.wikipedia.org/wiki/Action-sp…
You know it’s gonna be good when... twitter.com/getFreezeframe…
A good article on the acquisition of tacit knowledge —
Why Tacit Knowledge is More Important Than Deliberate Practice: commoncog.com/blog/tacit-kno… twitter.com/dubroy/status/…
Why Tacit Knowledge is More Important Than Deliberate Practice: commoncog.com/blog/tacit-kno… twitter.com/dubroy/status/…
Making pizza dough from scratch has changed the way I think about learning, knowledge, and what it means to "know something". A thread:
Love this. Same energy as "The C language is purely functional" — conal.net/blog/posts/the… twitter.com/8051Enthusiast…
"You have to poke things and see what happens".
Every now and then I relearn this lesson. A few hours of messing around can blow away days or months of "just thinking about it". twitter.com/Meaningness/st…
Every now and then I relearn this lesson. A few hours of messing around can blow away days or months of "just thinking about it". twitter.com/Meaningness/st…
Another rad project built with Ohm ☺️
Slang is an audio programming language by @kylestetz — try it out at slang.kylestetz.com.
Slang is an audio programming language by @kylestetz — try it out at slang.kylestetz.com.
Cool that CPython is moving to a packrat parser that handles left recursion in the OMeta / Ohm style.
PEP 617: python.org/dev/peps/pep-0…
PEP 617: python.org/dev/peps/pep-0…
Just discovered @gvanrossum wrote a great series of posts about PEG parsing back in 2019: medium.com/@gvanrossum_83…
@gvanrossum You might be interested in this: twitter.com/dubroy/status/…
📝New blog post —
Visualizing Packrat Parsing: dubroy.com/blog/visualizi…
All about the Ohm visualizer, and the design challenge of creating a full visual explanation of the parser's execution.
Visualizing Packrat Parsing: dubroy.com/blog/visualizi…
All about the Ohm visualizer, and the design challenge of creating a full visual explanation of the parser's execution.
📝New blog post —
Visualizing Packrat Parsing: dubroy.com/blog/visualizi…
All about the Ohm visualizer, and the design challenge of creating a full visual explanation of the parser's execution.
Visualizing Packrat Parsing: dubroy.com/blog/visualizi…
All about the Ohm visualizer, and the design challenge of creating a full visual explanation of the parser's execution.
Omar's work is consistently 🤯. Support one of the most creative minds in computing! twitter.com/rsnous/status/…
Thinking about hijack.social and @andy_matuschak's notion of _programmable attention_.
One observation: In a feed-mediated world, *we should all own our own algorithm*.
Apple could make this happen. When will iOS include a native feed that subsumes the others?
One observation: In a feed-mediated world, *we should all own our own algorithm*.
Apple could make this happen. When will iOS include a native feed that subsumes the others?
In the 2000s, security was a selling point for Apple (vs malware-invested Windows machines).
Now, it’s privacy.
Next, attention?
Now, it’s privacy.
Next, attention?
I don’t think the networks would expose their feed data. I’m thinking the meta-algorithm that determines where my attention goes.
“Instagram in the morning.” “No more than 15m per day on Twitter.” etc
mobile.twitter.com/borismus/statu…
“Instagram in the morning.” “No more than 15m per day on Twitter.” etc
mobile.twitter.com/borismus/statu…
Like an ad exchange, Instagram, Twitter, etc. would all bid for the next spot in my iOS feed.
Niche construction (en.wikipedia.org/wiki/Niche_con…) seems like a useful mental model.
If there's a Great Software Stagnation maybe it's partly because we've spent decades constructing a more & more specialized niche. That it turn constrains our thinking about what the design space is.
If there's a Great Software Stagnation maybe it's partly because we've spent decades constructing a more & more specialized niche. That it turn constrains our thinking about what the design space is.
The Internet Archive is awesome! I just discovered they host an emulated version of the Canon Cat: archive.org/details/canonc…
This was the big goal I had when I built the Ohm visualizer (ohmlang.github.io/editor/#30325d…): to visually represent the *entire* computation — failed branches and all — and to make it possible to jump to any node and step from there. twitter.com/geoffreylitt/s…
It could be more discoverable / learnable but I'm still pretty happy with how it turned out.
(I really should write this up somewhere.)
(I really should write this up somewhere.)
Showing a functional program as an icicle plot / flame graph is straightforward, but the challenge here was:
1️⃣ Ohm's parsing algorithm is not purely functional (due to how it handles left recursion), and
2️⃣ To represent both the parse tree and the computation in the same way
1️⃣ Ohm's parsing algorithm is not purely functional (due to how it handles left recursion), and
2️⃣ To represent both the parse tree and the computation in the same way
I am building a spreadsheet, please send help
5yo just now, pointing at her picture: "You see that? It's so small cause it's really far away."
Are there any good books/papers out there about how children learn *visual* language? It's fascinating to watch it happen.
Are there any good books/papers out there about how children learn *visual* language? It's fascinating to watch it happen.
TIL the Atari 2600 had collision detection registers: atariarchives.org/pmgraphics/cha…. An interesting thing is that the game logic can depend on the sprite colors.
Reminds me a bit of the "color sees" functionality in Etoys:
Reminds me a bit of the "color sees" functionality in Etoys:
This alarm clock is very @rsnous:
"The further the user moves or throws the alarm ball from the base station, the louder, the more aggressive and the more insistent the waking sound will be in the morning."
From "But how, Donald, tell us how?" (2002) dubroy.com/refs/But_How_D…
"The further the user moves or throws the alarm ball from the base station, the louder, the more aggressive and the more insistent the waking sound will be in the morning."
From "But how, Donald, tell us how?" (2002) dubroy.com/refs/But_How_D…
A cool end-user programming environment for a somewhat surprising domain: air traffic control 😬
Vizir: A Domain-Specific Graphical Language for Authoring and Operating Airport Automations — youtube.com/watch?v=rfmK4G…
📄 Paper: hal.archives-ouvertes.fr/hal-01886335/d… (from UIST '18)
Vizir: A Domain-Specific Graphical Language for Authoring and Operating Airport Automations — youtube.com/watch?v=rfmK4G…
📄 Paper: hal.archives-ouvertes.fr/hal-01886335/d… (from UIST '18)
I love systems like this that solve a specific problem in a very specialized domain. I think many end-user programming systems fall into the trap of trying to retain all the power of general-purpose programming languages.
Making pizza dough from scratch has changed the way I think about learning, knowledge, and what it means to "know something". A thread:
2/ For the past few years, I've made pizza for my family almost every Friday night. I make the dough from scratch, and over the years I've given the recipe to a quite a few friends.
Inevitably, they say something like, "It turned out okay, but it wasn't the same as yours."
Inevitably, they say something like, "It turned out okay, but it wasn't the same as yours."
3/ The first few times I made the dough, I also wasn't thrilled with how it turned out. I had to make it at least 5-10 times before I really felt like I knew what I was doing.
That's because making pizza dough (and bread-making in general) involves a lot of *tacit knowledge*.
That's because making pizza dough (and bread-making in general) involves a lot of *tacit knowledge*.
4/ Tacit knowledge is "know-how". It's all the stuff that you can't — or can't easily — express verbally (or write down in a recipe).
Drawing a portrait, playing piano, throwing a football: these all involve tons of tacit knowledge. Reading about them doesn't get you very far.
Drawing a portrait, playing piano, throwing a football: these all involve tons of tacit knowledge. Reading about them doesn't get you very far.
5/ Even things that are considered "knowledge work" — like being a programmer, research scientist, or CEO — require huge amounts of tacit knowledge. It goes by different names: skill, experience, perception, intuition. Or we just call it *talent*.
6/ Ultimately, what matters to me is being _skilled at X_ — and not just _knowledgable in X_. The pizza dough is a vivid reminder that there's a big difference between the two.
So what does this mean for learning?
So what does this mean for learning?
7/ When we think about learning, we usually focus on explicit knowledge, aka "the content". We look for ways to make it more relevant, relatable, interesting, and memorable.
This almost a kind of salience bias: we focus on the most visible/legible aspects of learning.
This almost a kind of salience bias: we focus on the most visible/legible aspects of learning.
8/ So one interesting question is: what kinds of tools and structures can improve our ability to acquire know-how?
@andymatuschak has thought a lot about this; he calls them _enabling environments_ (notes.andymatuschak.org/z244xx3kMf1v8U…).
@andymatuschak has thought a lot about this; he calls them _enabling environments_ (notes.andymatuschak.org/z244xx3kMf1v8U…).
10/ This makes me wonder: what are the essential features of learning-by-doing? To what degree are these features specific a particular domain (e.g., music performance, sports)? Are there patterns for learning more abstract concepts in the same kind of way?
11/ The concept of "deliberate practice" (Ericsson 1993) is relevant here, but afaik most of that research focuses on things like music performance and sports.
It's not clear to me how much of that applies to different kinds of learning, e.g. learning a programming language.
It's not clear to me how much of that applies to different kinds of learning, e.g. learning a programming language.
12/ But, I suspect that some things are similar. E.g., the importance of informative feedback on the results of the performance.
Interestingly, in baking most of the feedback is not immediate. It's hard to judge until it comes out of the oven.
Interestingly, in baking most of the feedback is not immediate. It's hard to judge until it comes out of the oven.
13/ Also, deliberate practice as studied by Ericsson typically involves expert instruction, a known "best method". In many learning scenarios, we have neither.
Tweeting As A Subversive Activity twitter.com/jasonyuandesig…
The real innovator's dilemma.
aka build something cool, or build something useful
From "The New Scholarship Requires a New Epistemology" by Donald Schön: neillthew.typepad.com/files/schon-ne….