Read on Twitter
Ok starting to kick the tires on @RoamResearch ... this is going to be my starting point. https://twitter.com/adam_keesling/status/1196864450520727552
So it's like workflowy but with transclusion links? I like the premise so far. I've used workflowy half-assedly for years and this zettlekasten thing sounds similar to how I do index cards though way too disciplined.
Not sure I like bulleted lists as the first class citizen. My basic unit is the paragraph and I don't think in lists well. We'll see. There's a soft line break within a bullet point, but no way to just have an unbulleted prose chunk as a page it seems.
What's with this Daily Note thing... is it a soft prompt to create a journal-like log as a throughline for non-chronological notes?
Slightly unclear on how this is not a wiki.
Slightly unclear on how this is not a wiki.
My first expt is to try and bring notes from Scrivener for my book project over. I have a folder called "nuggets" that I'm porting manually. Copying smaller fragments to roam and moving larger chunks to a different Scrivener folder. Roam is clearly not for composing long texts.
Wtf is a cose layout?
This thing needs Hebbian synapses. I'm creating a big pile of chaos here, in large part because most of the organization is latent, in rhymes and associations that don't have strong overlap in vocabulary. Still unlinked-mentions is a sort of ersatz Hebbian synapse mechanism
Ok I have my first feature request... email-based page creation. And some way to capture tweets and tweet threads in 1 step as pages.
My dangerous working theory is that if I dump all my unintegrated fragmentary thoughts, discovery notes, and links for book project into Roam, when I work on the chapters of the book, I can search this set of notes or wander the graph to collect up what I need for JIT compilation
So I can confirm one thing: the tool does encourage you to frictionlessly capture latent connections explicitly. I'm not seeing "new" connections as such, but finding it much easier to capture ones I've already seen cleanly.
Another thing it is forcing me to do is clean up my foundational definitions to avoid creating redundant pages that will create merge-work later. The merge feature is a powerful forcing function.
I'm creating a real mess here, but hopefully it will lead to enlightenment soon.
Hmm... I've reached a point of complexity, where it's actually helpful to create what I called a Dramatis Memetae. Something between an index and a glossary, but with a bit of structure to it. https://twitter.com/vgr/status/1189782933831180288?s=21
Yikes, the current view of my graph looks suspciously like a yak. I might have to shave this thing.
So far this is a view of a single project. I wonder if I put more disconnected projects in if it will start looking like a zoo.
So far this is a view of a single project. I wonder if I put more disconnected projects in if it will start looking like a zoo.
Alright, spent nearly 6 productive hours digesting 70% of email notes and 30% scrivener notes from multitemporallty into Roam, and found it improved greatly in the process. Still a condition of Chaos > Order, but Order is now winning.
To do:
* Rest of email and scrivener
* ~100 twitter bookmarks
* ~50 index cards
* ~50 iPhone photos of book pages
* ~12 scapple mind maps
* ~ half a moleskine equivalent paper notes
* ~1 moleskine equivalent iPad pencil notes
My second brain clearly has mad cow disease
* Rest of email and scrivener
* ~100 twitter bookmarks
* ~50 index cards
* ~50 iPhone photos of book pages
* ~12 scapple mind maps
* ~ half a moleskine equivalent paper notes
* ~1 moleskine equivalent iPad pencil notes
My second brain clearly has mad cow disease
This feels right because processing into Roam feels like progress against entropy, not merely porting from one unsatisfactory tool to another. Moving the ball forward rather than sideways. I’m gaining authoritah.
Tentatively, I think this deserves hype it’s getting. If they solve intake, it will be a genuine paradigm shift. They’ve managed to actually do what we tried valiantly to do with trailmeme at xerox a decade ago. I can spot all the things they’ve done right that we got wrong.
Still wrapping my head around transclusion as a working concept rather than vaporware ideal. It’s not been really necessary so far though I’ve done a few to get a feel for it. I suspect doing a second compression pass after this discovery/intake pass could be transclusion-heavy
The fact that I put in 6h of work despite intake automation being basically non-existent and high-friction suggests that once even a few basics are in place there (email-to-db would be my most used, followed by something to grab tweets easily) this could really take off.
I guess I leapfrogged past Notion, and there were a couple of other tools I started using like http://paper.supply but in view of this, I don’t think others hit the right mix. Hey @vidy___ you still working on your thing?
Interesting reflection. I wouldn’t have paid much attention to this 20 years ago. I had the heavy lift capacity to hold this kind of big-mass memeplex in my head. There is a sense in which this is a prosthetic particularly well suited for aging brains with big picture fatigue.
Fox/hedgehog may actually be patterns of aging, based on relative rates of decline of distinctionist, pointillist, phenomenological cognition vs integrative, holistic, reified cognition. I need my second brain to be a hedgehog. Wonder if hedgehogs could use roam as a fox brain.
Or possibly every tool has a foxy or hedgehoggy bias and this is the key stylistic difference between @fortelabs and @Conaw 
Roam as hedgehog second brain for natural foxes, Evernote as foxy second brain for natural hedgehogs
You two should do a podcast conversation together
Roam as hedgehog second brain for natural foxes, Evernote as foxy second brain for natural hedgehogs
You two should do a podcast conversation together
That whole idea of the power of centralization in a decentralizing world has 2 solutions: internal and external locus. Either your first brain centralized or your second brain is.
2x2: x = first brain, y = second brain both from centralized to decentralized
CFB/DSB: Hedgehog
DFB/CSB: Fox
CFB/CSB: Cactus
DFB/DSB: Weasel https://www.ribbonfarm.com/2014/02/20/the-cactus-and-the-weasel/
CFB/DSB: Hedgehog
DFB/CSB: Fox
CFB/CSB: Cactus
DFB/DSB: Weasel https://www.ribbonfarm.com/2014/02/20/the-cactus-and-the-weasel/
This probably also explains why @swardley and I beef over 2x2s and why I don’t really use wardley mapping myself despite recommending it enthusiastically to other people
Alright one week in, I've settled into some steady habits using roam, so there's a habit attractor here. Progress has slowed to more day-to-day but it still feels like winning the war on entropy as opposed to losing it in scrivener, email, or twitter.
Gonna start documenting subtle features here. First up: Roam is probably the most efficient prosthetic memory I’ve used, as in, creating efficient recall of thoughts I’ve already thunk and written up. A killer feature when you’ve written as much as I have (good or bad).
On a scale of 1 to 10, ranking media I’ve used
Paper: 1 (no recall aids unless you create ToCs or indices yourself)
Email newsletters: 2 (weak searchability, weak theme/thread continuity, weak gestalt, though substack is better than mailchimp)
...
Paper: 1 (no recall aids unless you create ToCs or indices yourself)
Email newsletters: 2 (weak searchability, weak theme/thread continuity, weak gestalt, though substack is better than mailchimp)
...
Twitter: 4 (strong searchability, strong threadability, weak gestalt)
Wiki: 5 (strong searchability, medium threadability, strong gestalt)
Blog: 6 (strong searchability, strong threadability, strong gestalt)
Roam: 8 (all of the above plus low friction update/create/rename)
Wiki: 5 (strong searchability, medium threadability, strong gestalt)
Blog: 6 (strong searchability, strong threadability, strong gestalt)
Roam: 8 (all of the above plus low friction update/create/rename)
The thing is, recall is a virtuous cycle. The better a medium supports recall, the easier it is to attach new information in the right places. Which makes recall even easier. And the easier it is to add content, the faster this process snowballs. So a compound interest effect.
Knowledge generally depreciates, and a holy grail of knowledge modeling and capture for a long time has been to reverse the default negative interest rate it accrues and turn it positive. There was a “knowledge banking” project at Xerox back in the day that aimed at this for eg.
In the past most such efforts have failed because they relied on automation to try and keep entropy at bay, which sort of works weakly with relative legible and structured information. But for more squishy information, the only thing that works is a “many eyeballs” process.
Anytime I’ve seen a positive-interest-rate knowledge repo, it’s because many people were resurfacing bits and pieces at random and making point improvements. So the only known solution to date for positive interest rate has been to collectivize and socialize the prosthetic memory
So a good measure of the technical sophistication of a medium is to measure the number of people (eyeballs) and money (how much they’re paid) required to sustain a positive interest rate. I think Roam reduces both to the limit: 1 person and self-funded hobby time.
Compare for instance Wikipedia (~100s-1000s unpaid) or Stanford Encyclopedia or Wolfram Mathworld (single digits -100s, but paid, directly it indirectly).
This is probably because Roam allows one mind to effectively act as many (Fox > Hedgehog)
This is probably because Roam allows one mind to effectively act as many (Fox > Hedgehog)
In practice, I’m finding now that when I have a new nugget to add, from whatever headspace I’m in (working, at the gym, random thought during in a meeting, half-asleep thought in bed), I can always make a quick judgment of approximately where a thought belongs, and put it there.
By contrast in weaker media, I have to be in high energy, direct focus headspace, having warmed up around 30 minutes to get situation awareness around the whole project. Only then can I reliably lower entropy and turn interest rate positive. Roam lowers threshold to 10% of that.
This btw is the maker time/manager time problem pg wrote about. Making needs 4 hour chunks because anything less tends to increase entropy rather than decrease it in any non-trivial knowledge work project. So anything that lowers that lower limit is a big win.
My suspicion is, a good KPI for a knowledge tool is minimum threshold of time required to make a negentropic update to it, with every halving of the threshold increasing its capacity to hold positive-interest-rate knowledge repos by an order of magnitude.
So a tool with a 2h minimum can sustain a 10x bigger positive-rate knowledge base with the same budget of people/money than a 4h minimum. I suspect for Roam, a very suitable project may get it down to minutes, and for a typical project, maybe 30min.
Something like that.
Something like that.
Hmm trying a 2x2 :diagram in roam. It is pretty janky but close enough to usable that for simple ones, I'm likely to use native rather than import from a more complex tool.
One sign I'm achieving some sort of "hedgehog closure" is that I'll find a thought in my pre-Roam notes I'm not sure I've captured yet, and when I go to add it in a page I'm not sure exists yet, I'll find that the page and the capture both exist. Falling necessity of deduping.
Okay, I started using Roam on Nov 26, so I'm now almost 3 weeks in. I seem to be entering a more difficult middle phase where it's all about resolving mergeconflicts. Lots of page merges and clean-ups today.
I am experiencing a kind of grind that I'll call a "Windows 95 to Windows XT" experience. My current state is a complex thing (Windows) built on a primitive simple thing (DOS) and I have to now go back and rebuild foundations in light of the clarity achieved at the complex levels
This is definitely not as fun as the first 2 stages. I think I'm going through a double freytag process here.
1. Discovery (pre-Roam) increasing entropy collection
2. Sensemaking (stage 1 Roam)
3. Valley (mergeconflicts and rationalization)
1. Discovery (pre-Roam) increasing entropy collection
2. Sensemaking (stage 1 Roam)
3. Valley (mergeconflicts and rationalization)
Hmm. Conflicted about what to do about the content I started putting into @readingsupply a few months back. While Roam is clearly a much more powerful paradigm, I do like RS's cleaner presentation and collaboration/permissions model. https://twitter.com/vgr/status/1199511350919065600
(in the quoted earlier tweet in this thread, I mistakenly referred to it as paper dot supply)
I was briefly kicking the tires on that a few months back, but unfortunately, it didn't take as a habit the way Roam appears to be doing for me. https://twitter.com/vgr/status/1140787641106911233?s=20
I was briefly kicking the tires on that a few months back, but unfortunately, it didn't take as a habit the way Roam appears to be doing for me. https://twitter.com/vgr/status/1140787641106911233?s=20
It'd be nice if http://reading.supply or something like it were bolted onto Roam as a sort of classic publishing/sharing front-end.
I was briefly being pretty evangelical about reading supply, but the enthusiasm dropped rather quickly once I realized the creation workflow friction is too high... on par with say google docs. But still, there's some stuff there that's interestingly different from Roam.
Unlike Workflowy, which I think has been strictly superseded by Roam. Again a nice product, but simply beaten comprehensively by a better one. For now. I don't think it's down for the count yet. https://twitter.com/vgr/status/1207437689709969408
Important threshold: it’s now easier to create a page linked off the right existing page for a one-liner thought than to tweet it and then postprocess the tweet. Took some practice, but I do it routinely now. Roam has hacked the tweeting impulse. Like Uber hacked taxi-hailing.
When I compare to effort in late 90s (hand code new html page, edit existing to link, and ftp to home page site) the effort level to create a meaningfully linked page has fallen fallen faster than Moore’s law. I now do it on phone for 1-liners. Amazing.
Kinda like what git did to code commits and forks. Made it really cheap and got us to continuous deployment/continuous integration.
Roam is CD/CI for brain.
Roam is CD/CI for brain.
