Excerpts from a conversation about personal information management | Hacker News Reader

Excerpts from a conversation about personal information management

sachachua.com

211 pointsJNRowe81 comments

The note on semantic and operational distinction of notes is interesting. I personally ditched hierarchies when I switched to Org-Roam. I used to think all the time where a specific note would belong - should I organize my notes by dates? Should I use the datetree feature of Org-Mode? Should I put everything in one file or split between multiple files grouping notes by some categories or tags.

These days, the only question I have to ask myself is "in what context do I want to rediscover this note?". For example, I don't usually sit around thinking: "Didn't we discuss this SSH-related problem with Jeffrey and Anna back in May? Let me go to the may-2024 folder of my notes and grep through them...". Instead, I would just go to either of these notes titled: 'ssh' or 'Jeffrey' or 'Anna' and search for backlinks, where I will surely find my notes related to that discussion, even if they're spread out across multiple days and many notes in multiple places. And it doesn't really matter where specific notes are - which file, what nested hierarchy of headings, etc.

Zettelkasten really does work. You just need a quick an easy way of cross-linking different notes. I highly recommend this little book called 'How to Take Smart Notes', it's fairly small, you can go through it within an hour or so. And remember the famous quote of Richard Feynman: "Notes aren't a record of my thinking process. They are my thinking process"... If you don't find a good way of taking notes, you won't be doing a good job of thinking.

The point about linting still stands, though. I recently had to grep for a note I was really sure was there, but didn't show in the roam-find- autocomplete. Turns out, at some point I must've accidentally put a stray character that made properties section stop parsing, dropping the note from the system. Another case is when I occasionally add what I intend to be roam note as a new org file, and then forget to press the shortcut I have to give the file an ID.

(EDIT: Similarly, over the years I had a few cases of some TODOs I forgot about because I accidentally made a whole subtree stop parsing with a stray character. Rare as it is, I'm beginning to wonder if I shouldn't switch to modal editing a la vim, as those mistakes tend to happen when moving through the outline with "speed keys".)

In general, Org works well for me, but damn if the fragility of plaintext doesn't bite me every now and then.

> fragility of plaintext

For that, I have .dir-locals.el file in the root of my notes folder with a single line: ((org-mode . ((eval git-auto-commit-mode 1))))

Even if I accidentally make a change, there's always trackable history

I'm more of a "stick everything in Dropbox/OneNote" kind of person myself, as it gives seamless syncing of my Org files between multiple devices. For various reasons I've been relying on that much less recently, so I'll reconsider the git approach.

Yes, I do the same, I use Resilio. I thought about switching to Syncthing, but my NAS supports Resilio out of the box, so I kept using it. I let my .git folder to be synced between devices and I never actually pushed anything to a "proper" git forge - I can't think of a practical use case for pushing my notes to GitHub/Gitlab.

My git-autocommit technique only for tracking unforeseen changes - I was using Orgzly on Android, and one day I tried using its sync feature and it borked up a bunch of my notes, creating duplicate files, etc. I didn't like that.

> The point about linting still stands, though.

Do you run org-lint on the save hook? I wonder if that'd be too distracting, or if it can be done silently unless some errors detected.

No, but this submission and thread just convinced me to try.

You can use a combination of auto-save-visited-mode (this will save to the actual file, rather than a backup) and gitwatch to auto commit all changes. It should prevent most forms data loss or desynchronization.

Never heard about it. I use nvpy, a simplenote clone.

How can I install org-roam? Looks difficult.

Org-roam is an extension that works on top of Org-mode, and it's very simple to switch to if you already know Org-mode.

Org-mode by itself, if you've never used Emacs before, can certainly be challenging - it has too many features. One of the best overviews of the Org-mode's built-in features is Rainer König's playlist on YouTube: https://www.youtube.com/playlist?list=PLVtKhBrRV_ZkPnBtt_TD1... He also has a Udemy course https://www.udemy.com/course/getting-yourself-organized-with..., but I've never tried it myself - by the time he published it, I was already well-versed in Org-mode features.

Besides these, Org-mode is extendable with other packages. I, for example do my pdf-annotations, presentations, anki-card editing, http endpoints testing, dotfile management and many other things in Org-mode.

I have a flat directory of notes that I just grep through. turns out the prefect system I spent my twenties chasing didn't need to exist.

Wiegley's "Today's agenda has 133 items on it," now joins David Foster Wallace's "I received 500,000 discrete bits of information today," in quotes I wish I could recite to others to express how I think and feel.

Tangentially,

> I have over 30,000 tasks in my Org Mode overall. 23,000 of them are TODOs. Several thousand of them are still currently open. I'm never gonna see them all. Even if I wanted to, I'm never gonna see them all.

As a systemiser using GTD I have an ever growing list of items to get to should I complete items for the day, but I don't understand how you could expect to address 133 items in a day.

If the approach is to let low priority items roll over, that just seems like a recipe for dropping the ball.

Are people getting through 133 items in a day? That's 216 seconds per item.

Yes but I use OmniFocus for habit learning or reminding myself to take eg meds. Taking 7 meds or supplements in 20 seconds is doable :)

This approach can also be useful for researching routines or habits after a trip.

I don’t follow - what do you mean by this?

I haven't given the article the dedicated read that I intend to yet but my impression is that he does not expect to address all 133 items and it does not matter.

As someone who has compulsively accumulated lengthy to-do lists and buckled beneath the phantom of hopes and intentions of varying importance and ambition I find it enlightening that the possibility of completing a task other than those that are important enough to not have to write down to begin with can be measured by their meaningfulness and address according to this measure.

I gather that the .0050% of information that I consume daily is what is of the greatest priority and the remaining 99.995% is synthesized and iterated over the next day until it reemerges as something important. I suspect that Wiegley completes on average about 5 important things each day. This sounds like a solid baseline.

It entirely depends on which tasks you have and where they are coming from. Not every task has to some unique hour-long work. People who are using routines, pre-defined lists or elaborated project-planning, usually have very detailed and long daily lists full of small tasks, each in the range of some second to minutes of work.

Maybe the 133 items are 120 items of one minute work each, and the rest are conditional or optional task you will not do that day.

> Maybe the 133 items are 120 items of one minute work each, and the rest are conditional or optional task you will not do that day.

Then stop managing them with fancy computers and just do them.

It may be helpful to have future tasks written down even if will only take a few minutes, something like “remind someone about something” and recurrent stuff. Then there are checklists.

If it's the Org Mode agenda view in anything close to its default configuration, those 133 items will contain not just tasks scheduled for today, but also any incomplete tasks that were either scheduled for time before today, are past deadline, or have a deadline coming in the next 7 days (IIRC, maybe it's 14).

This is to say, if you don't keep on top of current work, the agenda view quickly turns into an ever growing wall of shame.

I have a pretty weird love-hate relationship with it because of that.

I don't understand that, at some point don't you just forfeit what you haven't done in a month or two and get back onto a manageable list? I mean keeping track of everything is a philosophy, but to the point where the vast majority of your system is just noise, what's the point?

> Noise is either a sound of too short a duration to be determined, like the report of a cannon; or else it is a confused mixture of many discordant sounds, like the rolling of thunder or the noise of the waves. Nevertheless, the difference between sound and noise is by no means precise. — Ganot

http://www.websters1913.com/words/Noise

> Ganon

I was wondering what an action-adventure villain was doing in the dictionary

> Ganot

Oh, that sounds more like a real person.

Thank you.

Perhaps my locating of the quote is redeeming

https://archive.org/details/elementreatisephys00ganorich/pag....

> Ganon

For a second I misread it as Gowron, and my mind started replaying the quote above in an angry Klingon voice.

This is why I don't let my to-do lists automatically roll over. If it wasn't important enough to copy to today's to-do list, then maybe it's okay if it's never done. But I do keep them around in some form, because maybe the old to-do items had some interesting ideas associated with them that I might want to re-visit later. Or keep the 'cool ideas' and 'tasks to take action on said cool ideas' separate.

https://www.nuke24.net/docs/2024/202410-to-do-lists.html

Noise is good. There's nothing wrong with noise. Internet is noise. You only need a good search engine or discovering technique. For Org users there are plenty of different choices - Org-Roam and org-roam-ui; built-in org-agenda, tags and todo stickers; Denote; Khoj, plain grepping, etc.

Noise is bad. Horrifyingly and anxiety inducing. The Internet may be noise, but you never actually see it directly. If one could, it would likely drive them insane. Search engines and other discovery methods is how we view it, and why it works at all.

That's what I'm saying. I don't see all my notes at the same time; I never try to see the entire picture nor do I ever worry about composing a new note, thinking it would just get lost in my mess. My knowledge graph looks like a complex web of interconnected nodes. I have over two thousand indexed notes and a few thousand plain-text, unindexed ones. The tools I've listed above do help you to find what you need easily. I treat my notes as my "personalized Wikipedia" - I don't need to know how it is organized or structured, where specific things physically exist, or in which file a specific text of a note sits. It is noise, but it is useful noise, because I have instruments to "extract music" out of it.

At record time, don't know what is signal and what is noise. Separation of signal from noise is not persistence problem, it is presentation problem. Model is sophisticated: number of delays against action intended time, etc. flow into whether task is noise. Being in list is protection against amnesia but not protection against prioritization.

Indeed. But the action of creating is a list is separate and indeed different from actionioning said list. Creating lists reduces my anxiety significantly...

It used to reduce anxiety for me too, until my brain learned that, unless acted upon immediately, those lists get out of date within couple days, at which point they never shrink. More than once I ended up with lists that rot faster than I'm able to fix them, so I eventually stopped planning too much in advance.

Why not reduce the scope of your future plans then?

I dont think it's as simple as that, after all there are book written about task management.

When "something" comes up, it's not always clear to me whether it's urgent or not, important or not. Getting it out into a list frees my mental capacity.

For example, I got a reminder to renew my passport yesterday. Not urgent, but important. Goes onto a list. I'll eventually prioritize and schedule it in my weekly review.

The key (for me) it aggressively devote attention to pruning these lists.

Breadth is not the issue, depth is. Nor there is much that I can reduce horizontally. At work, two or three open tickets + some blocked work in the "back burner" can already trigger the problem; on personal side, there's even more stuff that I need to keep track of, and take care of in parallel.

Those few "active" items and some more "standby" items are easy to list and read by themselves, but unexpanded, they're also non-actionable. And what I learned is, trying to expand even a few of them more than one level down quickly becomes overwhelming, and creates big overhead - as I work on one thing, the task breakdowns for the rest go stale, requiring additional effort to fix them. It surprised me just how fast this happens (and how quickly it leads me to stop looking at my own plans).

Related and perhaps extreme example that taught me much is when I took a medium-sized ticket that seemed perfectly doable in two to three work-weeks, and attempted to break it down all the way to actionable TODOs no larger than 2 to 4 hours worth of work. I wanted to see if this would streamline my work and allow me to make a more precise time estimate for the whole ticket.

I probably spent a day or two on the breakdown itself, complete with estimates and dependency links for every item - starting Z requires X and Y to be done, Y is done when A and B are done, etc. Standard project management stuff - lets you compute critical path and prioritize accordingly, and even draw a Gantt chart. My 3-week project ended up having some 150 tasks in it. Initially it looked great, but just a day or two into actual work, I found myself redoing large parts of the breakdown. Every two or three ticked-off tasks, the newly-gained knowledge made it apparent some task dependencies were to too strict, or entirely unnecessary. Large subtrees had to be split, shifted around or deleted, or had to have their estimates adjusted, all while new actionable tasks had to be added (and broken down). And that doesn't even includes the time spent readjusting the list after being pre-empted by some unrelated, critical-priority tasks.

All that planning quickly became a huge maintenance overhead. I only ever started making progress on the project itself when I stopped paying attention to my lists. That experience has taught me to stop breakdowns much earlier, and to stop eagerly breaking everything down to the same level - however good estimates I got this way, they weren't even needed for anything, and at the same time, they were constantly invalidated by added re-planning overhead.

These days, I don't break down work ahead of time beyond what's apparent and useful at the moment. I went from ahead-of-time to just-in-time.

Another factor is, it seems I'm really bad at the whole separation of planning and execution stages. My mind doesn't work that way, and can't stay in "execute this list one item at a time" mode for very long.

So then why not change into a job that requires less mental effort? Or reduce your expectations outside of work?

Makes no sense to torture yourself over something you can literally change tomorrow.

Well, I did figure out the solution to my own problem. Also, I mostly like my line of work.

Also, "something you can literally change tomorrow" is very much wrong, for the usual obvious reasons: I couldn't possibly afford a 90% salary cut, and neither could my family. Given standard life choices, past a certain age your career becomes a shackle.

‘can’ doesn’t equal you likely would want to endure the pain of doing so, or that it would provide an overall net benefit… I chose that specific word for a reason.

Of course there are significant tradeoffs to be made, since there are infinitely many potential tradeoffs in life.

What an incredibly on-the-nose anecdote, I love it!

The term of art for this strategy is "size to the horizon". Imagine you're looking across an open plain. The trees and rocks closer to you are bigger and you can make out more detail. The ones further away are still abstract.

You have to know exactly what to do with the things right in front of you, but you have to keep only a general awareness of that which is distant.

Plain text files are cheap and occupy basically 0 space. Whenever I have an overfull todo list - let's call it todo.txt - I copy it to a file named x.txt, and trim it down. Then when I finish that, I go back to whatever is left in todo.txt.

Now pretend that I cut and cut but can't bring myself to reduce x.txt to less than, say, 50 items, every one essential to complete by today. What do I do then? I copy x.txt to y.txt, and reduce x.txt to just what I plan to do for the next 4 hours. If that's still too long, I copy y.txt to z.txt, x.txt to y.txt, and reduce x.txt again. You could always start lower in the alphabet (a.txt) if you want more "space".

You get the idea. The point is, with text files, if it's got too much in it, create a backup version of that file and cut it down to size. Repeat as necessary until your todo list is manageably long.

That sounds like spending far too long managing TODOs, instead of DOing them.

Anecdotally, managing todos usually takes a few minutes, as I’m tracking them already. After that I have the whole day mapped out so I can work without being distracted by worrying about priorities. And having actual breaks everything is noted down, not shuffling around in your brain.

That's like telling the architect he's spending too much time thinking about building the house and not enough time building it.

I think this is interesting but I'm totally out of the loop.

He links to https://github.com/brabalan/org-review - what is that? What is org mode and org mode review? What is a sketchnote and how do you create one?

[Org mode](https://orgmode.org/) is a popular emacs plugin for organizing notes and todos. I don’t use it so I can’t be sure, but the reference to review is [probably this](https://orgmode.org/manual/Stuck-projects.html) about doing a weekly review of one’s tasks and projects, probably in line with [the practice](https://gettingthingsdone.com/wp-content/uploads/2014/10/Wee...) from the Getting Things Done productivity system. [Sketchnotes](https://rohdesign.com/sketchnotes) are an artistic visual summary.

HN does not render markdown

org-mode is an extension for emacs. It's centered around a plaintext markup-format which is specialized for organization and note-taking, hence the name. It has several other extensions build up on the core-extension, and it's format, many coming from the original project itself. And it has later branched out into other areas, like literate programming, personal databases, etc.

Seems org-review is one such extension.

And a sketchnote is a visual note, something where you draw/sketch your information. Basically a more freestyle diagram or something to tell a short story.

Just forget stuff like a cool person

I'm storing this quote into my second brain right now. I'm heisenseriousjoking. Filing something away makes forgetting very cool indeed; I know I can look it up later.

Do you then actually look it up at the right time? I often don't. My anxiety increases with the number of uncompleted items on the list, until it eventually stops me from looking at the list ever again. Filing something away becomes a license to forget.

Ah, well, isn’t a “to do” item, so no anxiety needed.

I’ve collected quotes for years: pretty fun to review from time to time.

If your todo list is becoming too long that’s when you need to break it by project and calendar frames (today, tomorrow, this week, next week, this month, next months). It’s a presentation problem, that you can solve with a task manager or a planner.

My giant pile is my inbox. I refile each task in one of the above boxes when I’m reviewing it. And when I’m doing, only today boxes is important (up 10 items where only two only 3 or 5 are inportant)

It seems this is excerpts from a video or screenshare. Does anyone have a link for that? I’m curious to see some of the things the speaker refers to.

Does anyone handle org-mode for attachments + mobile sync?

I use Autosync on android and a script that runs on my laptop. This works great for the text files, but the attachments/data files (say like screenshots) are a pain to find and open on mobile. Second, they're even more painful to add from mobile.

I'm fine with quite elaborate set ups if it solves this problem.

Not that I am aware of. The mobile story with emacs is pretty painful. I use dropbox + beorg (iOS). If I have to deal with attachments off spelunking through the dropbox folders I go.

Realistically, I dont see the mobile experience improving much. Given Emacs' utility though I dont mind it.

While the authors both use org mode, this includes a broader discussion that includes various ideas and tips, including the books "Atomic Habits", "Getting Things Done", and "Building a Second Brain". (Updated, since my previous comment thought it was only about org mode).

org-review is interesting, but I just added another TODO state: DEFER

Then projects, todos, agenda items etc can go from TODO -> DEFER and I know that they're "not now" items. That has seemed sufficient for me. Tracking exactly when they're reviewed has been too much, and not everything needs a scheduled time in the future for review.

The approach in the article by strictly controlling everyone with AI sounds absolutely dystopian. I wouldn't want to work in such a company.

This is an excellent, if meandering, discussion -- lots of great nuggets in there I want to follow up on. Thanks for sharing!

I was hoping there is an audio/podcast version of this, but I haven't found it yet.

This is a satire ... right?

Haha, no.

But I know the feeling. I've seen videos about people who have special physical filing cabinets for notecards that take up an entire tabletop. They have to figure out how to number these things, which can open up a can of worms and lots of differing opinions.

This can seem farcical at times. Sometimes the knowledge-management world can seem like a manifestation of OCD or perfectionism.

But to be clear, I don't want to discount in any way that such approaches could work for many people on many projects. My general take is that if a person is being mindful about the _effort in_ versus _benefit out_, they'll probably end up in a pretty good place.

For YouTubers in a niche of a niche, sometimes there is a positive feedback loop to just go deeper down the rabbit hole. For example, once a content creator has "committed" to a paper-only Zettelkasten system, what are the chances they are going to "mellow out" and move to a hybrid paper+digital system?

For a YouTuber it's more advantageous to keep switching between various systems to keep generating more YouTube content.

> special physical filing cabinets for notecards

It sounds like you're a bit dismissive of it. The creator of this system Luhmann was a very prolific writer [0] which he credited to this system.

0: https://en.wikipedia.org/wiki/Niklas_Luhmann#Note-taking_sys...

> For a YouTuber it's more advantageous to keep switching between various systems to keep generating more YouTube content.

That's a factor, yes, but before I will grant that it is more (or less) advantageous, we would need to have a much longer conversation. (Right now, however, I'm not particularly interested in running experiments and drawing conclusions about how to best optimize YouTube content creation.)

It would be too coarse-grained to say I'm categorically dismissive of paper-based systems. My views are more nuanced.

Right, I'm aware of Luhmann and Zettelkasten and ZK-inspired systems. I'm glad it works for him and many others.

Still, if one goes down the rabbit holes of ZK and knowledge management systems, you will sooner or later find some videos that leave the realms of productivity and venture into the farcical.

This is important to recognize: "There is a common mistake people [make] when [using] PARA, Zettelkasten, GTD, Bullet Journals [...]. / They assume each system is universal. / These systems were [NOT] built for mass consumption." - 2022 blog post by Zain Rizvi titled "PARA vs. Zettelkasten: The false binary" https://www.zainrizvi.io/blog/para-vs-zettelkasten-the-false...

Right, I may have read your comment as generally dismissive of personal knowledge management systems in general, but I see you have a much more nuanced approach. Thanks for clarifying!

So. Many. Tradeoffs. :)

Both John Wiegley and Adam Porter are insanely productive, so something must be working for them.

Underemployment drives active github profiles.

How do you know this? What do you mean? They are in the top X% of software engineers and computer scientists in terms of ... (some kind of productivity)?

I'm not familiar with Porter, but Wiegley is a heavy contributor to a number of open source projects, is the author of the Ledger plain text accounting system, has been emacs maintainer since 2015, operates his own software consultancy, and is very active in several open source discussion lists. The guy is a dynamo.

Perhaps I should have said "impactful" instead of "productive", but in my experience the two tend to be correlated.

John Wiegley has authored a number of highly used open source tools and libraries. So has Adam Porter (although perhaps not as much as John).

I wasn't comparing them globally with all SW engineers. My point was that they likely have accomplished more than most, if not all, the commenters in this submission. I, personally, would be cautious in criticizing their workflow unless I can demonstrate something better. Otherwise it's just armchair criticism.

Many of these are points I would agree with, except:

> [comparisons to] commenters in this submission

I wouldn’t make claims as to the level of accomplishment here, not least of which because I’m not willing to pick any one definition of accomplishment.

For example, I’ll avoid the temptation to conflate visibility and popularity with productivity.

> be cautious in criticizing their workflow unless I can demonstrate something better

Yes, making assessments based on comparisons against alternatives is important. And I’m inclined to give their approach attention and consideration, especially for people with their mindset and skillset. But generally, I’m not willing to grant any particular level of broader applicability for a given audience.

So I would reframe this discussion as a question: “Which of their practices are likely to work for [me/my team]?”

Like so many things in life, valid generalization is hard. Especially hard when no one knows the correct answer, so instead we assess better and worse. So, reinforcement learning. And RL in complex environments with large action spaces can be very data intensive!

  >This can seem farcical at times. Sometimes the knowledge-management world can seem like a manifestation of OCD or perfectionism.

  >But to be clear, I don't want to discount in any way that such approaches could work for many people on many projects. My general take is that if a person is being mindful about the _effort in_ versus _benefit out_, they'll probably end up in a pretty good place.


  >My general take is that if a person is being mindful about the _effort in_ versus _benefit out_, they'll probably end up in a pretty good place.

What point(s) of that discussion feel to you like satire? I'm not trying to be snarky; I'm genuinely curious. I remember myself when I was younger; I never viewed note-taking as a serious endeavor. Oh boy, how wrong I was. I wish I had found better ways of taking notes sooner.

I am so here for this comment. I forget things. If I don't take notes then I have to spend time rediscovering things. Searching my own notes solves that problem for me. I wish I had started taking comprehensive notes sooner.

Notes are not only to help you retain information. They are also a great research instrument. Whenever I need to give a status report, I can quickly compose the status update or even a presentation based on past experiences.

Because I use Emacs for my note-taking, I can quickly browse relevant GitHub issues, PRs, Jira tickets, Emails, etc. - without ever having to open or search for anything in the browser.

I can run code snippets in different languages and send HTTP requests and analyze and even visualize the results. I can have tables with spreadsheet-like calculations. I can turn my notes into Anki cards and synchronize them so I can then review them on my phone - spaced repetition is a great tool for memorizing things. I can create PDF annotations - right there, inside my notes - PDF document would be displayed in the next pane. I can store links to specific places in our codebases and annotate code without having to change the code or add comments. I can type something like 'RFC 6962' and immediately start browsing the relevant document (again, right there where my notes are). I can type a question to an LLM and get the answer inlined. I can look up any Wikipedia terms or search on Google, DDG or GitHub - and again, without switching to the browser. I can translate sentences and whole paragraphs between languages.

And the greatest thing about this all? I don't have to pay a damn dime to anyone - these tools are free. And I don't have to worry about some company taking my data hostage. I don't have to worry about converting my notes to a meaningful format - it is all just plain text. I can easily convert my notes, analyze them, gather stats, encrypt them, send them to LLM engine for processing, force Emacs to read them to me out loud. The possibilities are virtually unlimited.

No, emacs and note-taking are forms of pointlessly recursive refinement. Think of sharpening a sword you've no intention to swing. Hold up a hand mirror to your vanity mirror. The resulting wormhole is what you get with emacs and note-taking.

> Think of sharpening a sword you've no intention to swing.

More like having every intention to achieve glorious victory in battle with your sword, but never getting to it as you keep finding ways to make the sword even sharper. That's legitimate trap, yes.

Most are so afraid of this trap, that they just live on the battlefield and keep fighting with their dull, wooden swords; when asked if they heard about sharpening - or steel - they mumble some quotes from Arthur C. Clarke's story, "Superiority". And it's true, many people fall into this trap. Some are never heard from again. But then others occasionally share their sword designs and sharpening tricks, which end up improving weapons for the whole army. Then there are those who pop in and out, tinkering in the field, upgrading their blades until they become rocket propelled chainsaws or other weird shit.

The truth is, if you're not satisfied with your Army Standard Issue Sword and want something better, you'll need to fall into this trap once or twice. That's how you learn how to upgrade your own gear.

Or, in short: exploration vs. exploitation tradeoff. If you want to get shit done, you need to find a good balance. Maximizing either of the two is a failure.

“Do not-doing and nothing is left undone.”

Reply on news.ycombinator.com