Most-Watched Software Engineering Talks of 2024 | Hacker News Reader

Most-Watched Software Engineering Talks of 2024

techtalksweekly.io

285 pointscempaka88 comments

I really wish there existed a central platform that tracks/links to all tech conference videos from past to present because YouTube just plain sucks for this and in general, to focus. They've had DECADES to polish this experience (in general for any educational content), at least under the premium tier and I've given up all hopes of them ever getting this right.

Off the top of my head, such a platform could:

    1. Encourage and foster great discussions (and support markdown!)
    2. Have a "community notes" kind of feature to highlight outdated info, errata, related talks etc.
    3. Expose brand new conferences and lesser known ones
    4. Allow people to share curated lists
    5. Likes/dislikes restored
    6. View count tracking from across multiple sources (some confs host on their site before YouTube upload)
    7. Support for interactive viewing of decks, source code etc.
    8. Speaker profiles (w/ verification)
    9. Automation to help conferences/speakers/etc. to submit/update content they own/keep things in sync

etc.

I don't think such a platform exists today but would love to be wrong.

this sounds like a lot of work for something this niche

I don't think this is about a niche but rather a broader need for lists, more specifically, integrated lists from users. Pinterest, for example, is built around lists, Amazon has wishlists, IMDb[1] provides movie rankings, etc.

From a data portability perspective, the real challenge is getting users to connect all these platforms to a kind of unified list, making it accessible for queries, machine learning, or other applications. This, in my view, is a traction problem rather than a technical one, independent of whether it justifies a business. FriendFeed[2] was a very interesting project where users could connect their profiles with many services and their followers saw their updates. It seems their Tornado web server keeps getting updated [3].

More than a decade ago, I explored some of these ideas [4][5], conceptualizing a directed acyclic graph (DAG) model where data operations could be performed at a global level, dynamically recalculating like a spreadsheet.

Going to more complex topics there are resources such as differential dataflows [6].

[1] My movies ranking: https://www.imdb.com/user/ur0601133/ratings/

[2] https://en.wikipedia.org/wiki/FriendFeed

[3] https://github.com/tornadoweb/tornado

[4] https://blog.databigbang.com/ideas-egont-a-web-orchestration...

[5] https://blog.databigbang.com/egont-part-ii/

[6] https://timelydataflow.github.io/differential-dataflow/intro...

I think this is a great idea for a startup.

As a scientist, I would definitely be a paying customer, and on top I could deduct that membership from my taxes or have my lab pay for it like we (like many others) are paying Overleaf just under $1000 a year only to have LaTeX run in a Web browser to collaborate on papers (now my most important online service/app).

We usually know where to find the talks and recordings for the conference venues of our own specialist areas, but if there was a general "ScienceTube"-type Web site, you could also check out talks from neighboring disciplines. God knows, this might be as useful/addictive as HN.

Should you want to launch this & need someone for the scientific advisory board, please get in touch.

EDIT: typo fixed

>I think this is a great idea for a startup

Ideally, this runs like wikipedia or Lichess, if I'm being honest. I don't think its a great business model for a VC backed startup[0]. Perhaps a more traditional business inception might work, if you are selling it to the big conference organizations / hosts and make it attractive enough for it to be their go to for posting videos and such.

I think it ends up imploding against the pressures that come with it with being a VC startup though.

[0]: I'm taking the term startup to mean shorthand for venture backed startup, as it is usually used on HN and other places.

It's much more like wikipedia than Lichess IMO.

The hard part with going completely free is building enough of a community of strong contributors that they are able to effectively curate the contents for quality's sake, rather than the obvious commercial and career incentives to advertise their own work (or sneak in less educational content). That could be very hard to bootstrap without doing a ton of the curation yourself initially, and it requires an ongoing commitment to manage the community once it is established, which is also time consuming.

> I'm taking the term startup to mean shorthand for venture backed startup, as it is usually used on HN and other places.

I don't agree with that characterization because there are plenty of viable and useful "start up" opportunities that require zero outside funding, or only enough to get started. Also, VC funding is literally just debt that can convert to equity (it's a loan that can have an extremely high effective interest rate if you're successful), so you actually want to avoid it as much as possible if you think you have a good shot at accomplishing your goals without it. I think a surprisingly large number of founders and people working in SV/startups don't fully understand that.

Network egress fees are going to kill you, unless you have sufficient capital to run your own DC

I've had the same idea a few years ago and started working on it, but never found the time to push it over the finish line.

> Likes/dislikes restored

Why? It leads to a lot of abuse.

What you see as abuse some would see as user agency. User agency is a very powerful thing, and it's shortsighted to squash it just because your users aren't using it strictly to enhance your content in the way you want.

It's especially shortsighted to squash it just because they used it on some of your advertisers' content and your advertisers yelled at you about it.

I think my point is that it requires a (perhaps opaque) algorithm to work well. A pure popularity metric never led to good quality in my experience.

So even if you have your up/down votes, expect that how videos are presented to you are not strongly reflective of the vote count.

There's a good directory of conferences at https://confs.tech/ and at https://dev.events/

The security community used to maintain one that was fairly active pre-covid.

https://github.com/PaulSec/awesome-sec-talks

Checkout https://dev.tube/ site. It is not as feature-rich as you said but better than nothing!

I know these are conferences that they track, but I happen to know the Scott's talk from FOSDEM titled "So You Think You Know Git?" is at 1.3m million views:

https://www.youtube.com/watch?v=aolI_Rz0ZqY

Really surprised at the lack of “big” names on the list. My gut reaction was “really, no PyCon?” But when I went to the PyCon channel, even the keynotes from last year barely cracked a thousand views.

> Really surprised at the lack of “big” names on the list.

I think it's pretty cool we're hearing from more voices though.

There is nothing interesting going on at pycon, usually. Most language-specific conferences tend to be pretty boring imho.

Depends on the language. Julia has a lot of cool stuff going on. And really small languages like Elm (when it was still being used) would probably involve a lot of discussion about programming language history and future considerations

I just checked and my PyCon keynote got 6,200 - pretty happy with that! https://www.youtube.com/watch?v=P1-KQZZarpc

I consider you a pretty big name in our little corner of the world, so that number seemed pretty low to me! Expected 5 figures at least!!

[flagged]

Is number of YouTube views the new metric for gauging how much “people care about learning to program”?

yes

The days of people entering the industry to make big money are probably behind us.

good

hmm sounds like your AIE keynote was bigger than PyCon :) https://www.youtube.com/watch?v=eTTMUWP5B0s

I watched a little bit of the filesystems in rust talk and while it was a fairly straightforward talk, I was shocked at the q&a session. Emotionally hyperbolic ranting, jumping to conclusions, and wild accusations. I had no idea the linux community was so childish and reactionary towards rust.

What do you mean? That wasn't a Q&A session, they interrupted the speaker and ate all the time arguing. The speaker didn't get to finish his presentation.

This makes me think, whoever wrote these top 100 talks has either questionable tastes or just did a simple query and filter (SELECT * FROM youtube.video WHERE tag = tech SORT BY youtube.video.views LIMIT 100).

It clearly says the most watched talks, so yes. The author mentions in the article "from the conferences I follow"

That's not a good metric. For every Linux kernel drama, you drop something more interesting, like a less watched PapersWeLove conference.

Wish Jon Blow would get back out there! He was asked on a recent stream and said he has turned down invites while trying to get Jai pushed out the door.

I really want to learn his methodology to writing software

His methodology is to put his hands on the keyboard, write a function, a struct, make it compile, make it produce the correct output, make it faster, make it use less memory...

Considering how much software he has actually put out... do you?

Check the Data Oriented Design talk: https://m.youtube.com/watch?v=rX0ItVEVjHc

Jai has been in development for a decade. I would appreciate if it were finally open-sourced.

Considering the fact that Jon despises open-source you might be waiting a long time.

Why does he despise open-source?

There are many rant videos available online - but in general the core of his software philosophy is based in minimal, clean, extremely optimized code. Open-source rarely results in any of those things. Jon is the last guy wandering in the desert, still devoted to concise code - I love him for it.

I feel like he's barking up the wrong tree, opening up his source code doesn't mean the code will get dirty. He can simply refuse any contributions.

He’s stated that he doesn’t think open source helps produce high quality software and attracts a bunch of people that are less interested in building software and more interested in building policy.

However, he’s stated many times that he will release his language compiler and modules openly when they are ready and that he believes he’ll have to open-source it at some point if he wants broad adoption.

I do think that the `SOA` struct modifier keyword is maybe the most interesting thing out of Jai. Other languages would do well to think about adopting a mechanism to support something like it.

My understanding is that it's been moved to a convenience macro and is no longer a syntax-level thing. The original videos where Blow talked about SOA at the language level is very old now, and the language has gone through a lot of changes.

Oh, then the language primer is also out of date.

for AI Engineering here is what was popular (30k-300k views):

- Jerry Liu on Agentic RAG: https://www.youtube.com/watch?v=zeAyuLc_f3Q

- Emil Eifrem on GraphRAG: https://www.youtube.com/watch?v=knDDGYHnnSI&t=2s

- Justine Tunney on LlamaFile: https://www.youtube.com/watch?v=-mRi-B3t6fA

- Daniel Han on Low Level Technicals of LLMs: https://www.youtube.com/watch?v=pRM_P6UfdIc&t=5094s

theres a long tail of others that didnt get the views but i consider quality. we also do AI Engineering Management talks now https://www.youtube.com/watch?v=uiq95JYpBGY&list=PLcfpQ4tk2k...

Does this cover only GOTO conferences? If so we need to update the title to avoid being clickbait.

It appears to contain a bunch of other conferences. NDC, ElixirConf, etc.

No, there's content from lots of different conferences (though, GOTO is there a lot -- maybe a good indicator of quality?)

popularity ≠ quality

for a Mr. Beast video, maybe that's true, but for a specific audience like "Developer Conference Talk" probably a decent proxy

rust vs javascript ?

Having watched a lot of the goto talks, I'd say that they're like a good tech conference: ~5-10% of the talks could have a major impact on your work/career, 20-30% of the talks are interesting and slightly useful, and the remaining talks are basically product pitches / sponsor slots, which can be useful but are often biased.

If I'm not asking for too much, I'd love to see an example of a talk you consider a member of the "could have a major impact on your work" bucket

https://www.techtalksweekly.io/p/tech-conferences

List of conferences the author covers

As others have noted, there are some problems with the methodology of how this list has been created.

First, the list of "almost every" Software Engineering conference around the World[1] used, is just 72 items long. This feels light, and some conferences I would expect to see are missing: Brighton Ruby is very popular in the UK, but possibly a bit niche, so fine; but, FOSDEM however, seems more of a glaring oversight, perhaps explained by the fact that most of those talks are published not through a single channel.

The author has a mechanism to add confs to this. We could try and help them with that.

Then, the list seems to have ordered talks published by each conference's YouTube channel, by total views.

When one of those channels posts a talk twice (for example, the Linux Foundation's keynote of Linus talking to Dirk Hohndel), you end up with split user counts which means its lower down the list than you might expect.

I also think this means we're now measuring "most watched" in a slightly weird way: is this now just a reflection of marketing and subscriber reach? GOTO has 1.04M subscribers, ReactConf just 28.5K - is it any surprise that the list has more talks with more views from one of those over the other? Who is to say which conf had the better or more interesting talks?

Other engagement metrics (thumbs up, percentage of subs, number of comments, how many inbound links, and so on), might provide a better number to proxy for "quality" or "best" than just views, although the author isn't making a claim for quality, just views, so perhaps I'm asking for something beyond what is reasonable or what was promised.

This is a good start of an idea though, it wouldn't take much to make it better, so best of luck to the author or those who want to steal the idea: I'd love to see a better list in the future, and perhaps over a longer period of time than one year, too.

[1] https://www.techtalksweekly.io/p/tech-conferences

It's also missing most (all?) of the more academic conferences.

You are nitpicking.

They just did an algorithm to list a bunch of popular/nice talks for software engineers. I'll watch a bunch of them.

I'd like to see my fellow hacker news members to point to good talks that aren't present in the list.

Agreed - as a concrete example of an omission, I gave a talk at Philly ETE 2024 whose view count would put it at #14 on this list:

https://youtu.be/vQPHtAxOZZw

I didn't expect so many C++ talks in this list. Not bad for a supposedly dying language :)

"Introduction to Wait-free Algorithms in C++ Programming" by Daniel Anderson (https://youtube.com/watch?v=kPh8pod0-gk) is a very good talk, btw!

Reports of C++ death are greatly exaggerated.

It's obviously in its last breath, there are only billions of lines left.

This sounds like survivor bias, fostered by all the "written in Rust" in project titles (HN is a major contributor!). C++ is very well alive, e.g. in graphics, game development, embedded systems, just to name a few. It's a nuanced language for sure, likely not the most user-friendly, but it's not going to die anytime soon.

My comment was very much tongue-in-cheek. Poe's law stroke again.

I don’t like watching videos (or rather can’t watch at night, but can read). To date I haven’t found an AI that can produce a good article from a video. No, not just transcribe the video but actually produce a quality article with images and stuff from the video. Like a human who is instructed as “watch this video and produce a very high quality article that talks about the things talked about in the video”.

Has anyone has any luck with this?

It would be awesome if I can give this link to AI that will produce a PDF, each video being a chapter in the PDF.

> would be awesome if I can give this link to AI that will produce a PDF, each video being a chapter in the PDF

If you've got is slides+YouTube, then https://notebooklm.google [0] might work wonderfully well. It does for me. Though, it more Q&A than an article with illustration.

[0] https://illuminate.google.com is its limited/audio-only alternative.

There was a post on HN that just got removed a minute ago about a Syrian tech conference. Inevitably someone asked if there were any women speakers of note. Out of curiosity I counted up the femme speakers in this list. By the time I finished counting the thread was deleted so I'm gonna post my findings here:

13 out of 100

There were a couple with multiple speakers but they were always all masc except for one that had like 10 speakers

Linus's talk is on there twice. If you add up the view counts, it should be much higher.

RIP strangeloop </3

So what (if anything) is taking strangeloops’ place?

Hopefully nothing. I’ve been to three and it’s a huge self congratulatory waste of time for everybody. Like most software conferences

Sturgeon's Law is fractal, I have also been to three and don't regret it. Your frame of mind shapes an experience as much as the actual experience.

I mean, I expect as time goes on the number of people who get the reference goes down. Things get diluted.

It slowly goes from a "Here's some very cool esoteric thing, that you can use" for people that write business systems in prolog, to "here's how to connect workaday language with workaday db in a slightly unique configuration".

edit: I looked at the videos. 10 years ago the talks were Rich Hickey and "Production prolog".

Did Douglas Hofstader ever speak at strangeloop? I know he doesn't actually like computers.

"have you tried rubbing a database on it" feels like a spiritual sibling.

718 videos https://www.youtube.com/@StrangeLoopConf

Thanks. I wasn't aware of this channel. :)

I got to see some of these in person, though at different conferences. If you're at all interested in Java, the talks by Venkat Subramaniam and Nicolai Parlog are worth a watch.

> If you're at all interested in Java

Is that even a thing in 2025? I thought the consensus was that Kotlin has a strictly better syntax than Java.

(I'm not even a Kotlin fan. I was pretty sad when Kotlin started taking Scala's market share...)

Kotlin fixes a lot of the verbosity, but its tooling support is shockingly bad compared to Java in VSCode and IntelliJ for me.

Why is it that every time the word Java comes up, some rando without any apparent insight into Java shows up to make weird Kotlin remarks.

1. There is no such thing as "strictly better syntax", that's impossible to objectivize.

2. Kotlin, while growing in popularity, doesn't even hit the 1% mark on the TIOBE index. It is currently less popular than Prolog. Java sits at 10%.

3. Don't yuck someone else's yum. It isn't constructive. It doesn't promote good discussion. And this just reeks of uninformed, subjective "java bad"-bandwagoning.

this is a list from their videos, not all videos. There's plenty of software engineering videos with 5x or 10x more views

Here's 850k views https://m.youtube.com/watch?v=SO83KQuuZvg

look up "how to build and LLM" and your find plenty of videos with more than a million views from 2024

Sebastian Lague is amazing I liked his 3D world package drop game video

Seems like this over-indexes on a small handful of conferences (I had never heard of GOTO before this). It would be interesting to better understand the methodology here, particularly around discovering talks on YouTube.

I don't see anything ever dethroning this tech talk from no 1 position

https://www.youtube.com/watch?v=y8OnoxKotPQ

The title should include 'from tech conferences'. Otherwise no Karpathy here is surprising!

Are speakers being paid these days ?

Regular speaker here (who is also on the list).

I was never payed for a talk, and I have very rarely heared of people being payed by the conference itself -- appart from some "rock-star" keynotes.

Generally talks are sponsored by the person you currently work for - i.e. they pay for travel, and sponsor your time. If you work for a vendor in the space, this runs under marketing expense. If you work for a "regular" company this runs under personal development, employer branding & industry exchange.

Good OP but even better posts in thread.

Reply on news.ycombinator.com