Vlad's Roam Garden

Powered by 🌱Roam Garden

How do I read things on the internet


I have a somewhat elaborate process for reading things that I find on the web. I've been inspired to share it because after many a long iteration it finally feels adequate!

Reading things on the web seems like it should be easy, and yet - I've been failing at it for years! 🙀

In this article I explore my current workflow and challenges that made it into what it is today.

Workflow

object links open inside object which is not ideal =\

{{TODO}} do an excalidraw diagram

loop on SRS

can include the links to blocks in the labels

but need to make sure R.G. exports them (link them from elsewhere as-well)

➡️

🔁

Transcribe into a podcast ➡️

Often this is sufficient to get what I want from the article 🔚

➡️

➡️➡️➡️➡️

The core goals for the workflow are:

prioritize things that I want to read

direct my attention to them over time

afford for multiple level of processing of information

audio

skim-read

deeper reading with highlighting, annotation & setting up Spaced Repetition review

Help with information retention over time

Create support structures around the process of engaging with information

--

fractal structure

similar level higher level and deeper processing

Discovery

This is part of the pipeline that received relatively less optimization attention, mostly by virtue of me suffering from abundance of content rather than scarcity. I include it primarily for completeness’ sake.

Some ways in which discovery is happening for me are as follows:

Follow-up from previous things I read

people often link related content

Ampie extension is helpful to discover the broader conversation about a given piece - which includes links to related content

Recommendations from friends - sent directly or via social media

Aggregator newsletters (Indie Hackers, https://rationalnewsletter.com/)

I like aggregator newsletters because they introduce an additional layer of curation over raw subscription streams like RSS.

I'm also subscribed to a few "normal" newsletters

Reading Inbox

When I encounter something that I think would be worth my attention - I add it to Instapaper

I don't actually use Instapaper for reading, but adding things into it serves as a trigger for automations that would


The first problem of reading things on the internet is that there are too many things out there one is tempted to read.

Even if you have a good curation process there is always too much content and too little time.

My first approach to managing the reading inbox was to keep things I want to read in open browser tabs forever and breathe a sigh of relief when my browser crashed and all the tabs disappeared

When I noticed that this process didn't actually achieve the goal of helping me to read things I wanted to read - I started pushing myself to add things to Pocket/Instapaper to have a clear backlog of things to read

Which put me in a situation when I had hundreds of articles in Instapaper instead of as persistently open tabs (somehow that only marginally impacted number of open tabs I had 😅)

The result wasn't amazing - I went from not reading things and having them eating into my attention to not reading them and forgetting about them.

Arguably it was an improvement, as attention is an important and scarce resource, but as the point of this workflow is to help me actually read things instead of collecting the things I wish I have read - it was a failure.

A better way to direct my limited attention was called for! And I found it in Spaced Repetition

When a piece is added to Instapaper - a Roam "block" for it is automatically created

It's tagged with to/read and configured to become an SRS Card

The best way I know of to direct my attention programmatically is Spaced Repetition - I use it extensively for inbox processing, engaging with content over time and developing habits.

How does this work:

When the item is originally added to Roam - it's scheduled for a review in one of the next few days

On the day of the review - I decide if I want to read it and if not:

I reschedule it further into the future

Or mark it as "done" if I'm not interested in the piece anymore

It proved to be a good match for reading inbox handling. The above process has the effect of:

keeping the things I want to read salient

sorting them by excitement - things that I'm repeatedly not excited to read end up scheduled exponentially further in the future

Listen to content in audio form first

For any new piece of content I want to engage with - listen to an audio version of it first

This is often sufficient to get what I was hoping for from a piece ✅

If not - it serves as a first-pass skim read before deeper engagement


This is one of the core pillars of my reading flow — I think reading things in audio form is underappreciated.

Audio form dramatically extends the range of environments and situations when it's convenient for you to read.

I listen to audiobooks, podcasts and TTS version of articles when I bike to places, do chores and sometimes even while taking a shower.

This allows me to read more - in fact it increases my reading throughput to a degree that I can first-pass read things faster than I find new things to read!

Read it once mindset

An important stepping stone to make audio form work well for me was overcoming “only read a given thing once” mindset.

What I mean by that is that when I originally started using TTS to read things - after listening to an article - I felt like "I read this, I'm done with it an and don't need to engage with it anymore".

And while it's actually true for many types of content (opinion pieces, news articles, fiction) - I found it that for deeper, more technical pieces - just listening to something once, often wasn't quite satisfactory. I wanted to highlight paragraphs, add notes, play with presented models.

As a consequence I was avoiding listening to all content as I had a vague sense of unease "but what if it's a piece I want to engage deeper with and by listening to it, I'd lose an opportunity to derive full benefit from it".

Eventually I realized that it was silly 🙃

My new process is:

listen to all the content that comes my way first

this is sufficient level of engagement for a large chunk of what I want to read

for things that need deeper engagement

put them on top of the queue of the to-read things

read them again (likely in text form this time), highlight and annotate them, play with models they present, find follow-up reading

Spaced Repetition reminds me to engage with the piece until I mark it as fully processed

Mechanics - I currently have my custom tts setup, but I have a hope of replacing it with Reader when they fix things

got a few interactions right that were missing in the tts of prev read it later apps


after listening to audio- mark as done or put on top of the queue

Support structures

dedicated reading time/space, tech that enables it

getting a tablet to read from/ipad

dedicated device,

with ux affordances

but also having a mindset that the device is for "serious reading"

had kindle, but found it frustrating esp for the web & pdf content

and so mostly haven't been using it

and reading from the phone/laptop both of which felt subpar

physical books have nice aesthetics, but overall unsatisfactory UX

include highlighting with Hypothisis, review in Readwise and reference in Roam

ipad highlighting hack

audio notes

support the habit of reading with Beeminder

talk about routines as something that actually worked better

Key improvements over time

Some changes in the workflow that were particularly useful for me (highlighting them here as I think their importance may be lost when considered as the part of the overall structure):

Things I'm still unhappy about

taking notes alongside reading

handwriting sucks

a lot of friction for getting a kb and typing things out

external kb?

voice keyboards kind of suck

i’ve seen Jacob Cole using it to a certain degree of success. It actually seems pretty good so I should definitely consider using on this more on iPad.

audionotes lack context

tts ux

rn something needs to be a proper article to be transcribed

Outlook for the future/ New things I'm exploring

Readwise Reader

decent tts experience (though buggy)

snooze instead of just "later"

generally one can hope to incorporate learnings from this wf into a specialized tool and reader has a potential to be that tool.

things that are missing

attention management, or a way to simulate it (can't do my roam integration well bc can't subscribe to new things in reader)

getting tts experience from "promising" to "good"

fix the multitude of bugs

add

ability to do tts offline/pre-fetch tts results

auto-skip-back after pause like in podcast/book apps


Addenda

Failed experiments

spritz speed reading

audio takes role of first pass

if there is a second pass -

copying things into Roam and reading them there

kindle

Interesting asides

seeing highlights in context when you revisit an article in the future

promnesia, readwise, hypothesis