mdp

whisper to the rescue - https://gist.github.com/mdp/68cd55abd73cd5dd569f62c96eb3fa83

You can burn this on top of the video with `ffmpeg -i video.mp4 -vf subtitles=video.srt -c:a copy video_subtitled.mp4`

I've been really happy with DuckDB, and I love that fact that it can operate directly on things like JSON an JSONL. It's become my data swiss army knife.

Exactly. I respect their decision to go closed source if that's what they need to do to make it a viable business, but just be honest about it. Don't make up some excuse around security and open source.

This is fair, although I ask for it to be dark themed to match what I think was the style of typing game I remember growing up with (it's been a while). Bumped up the font though.

Perfect example is my local coffee shop that is 100% on Instagram only.

They've done amazingly well on just Instagram with the groups they are targeting. I doubt that a website would have any impact on their business. In fact Instagram gives them something much easier, more visual, and with a built in social feed (no need to setup a mailing list, just use Instagram).

"But it's a walled garden..." - Most people don't really care. And also, it's a coffee shop. If Instagram shutdown, they'd be on the next platform in a week and rebuilding the same following.

It's annoying to people like me, but don't see it changing anytime soon, and I can't really blame the business.

TL;DR: I think you should still learn regex, even though AI has made it a "useless" skill

https://mdp.github.io/2026/03/17/the-kids-are-alright-and-th...

I've had a hard time trying to explain this to non-engineers, but whenever I talk to a fellow developer they're struck by the same feeling. We've lost that magic flow state when your heads down on a problem.

Not to say that it's all bad, there's something quite fun about coding with AI, but I do feel like we're losing something that was fundamentally part of being a developer. I'm late to less meetings though.

We're launching a pre-seed/seed funding program where you can apply directly and get a decision within a week of your first meeting. $1M in funding for 6-10%.

We're 1984 Ventures - we've invested in 100+ companies at the earliest stages, including numerous YC companies (PostHog, Postscript, Shelf Engine, Convex, [Alex.com](http://alex.com/), HyperDX, Tandem, Avallon, OrangeSlice, Terrakotta, Olive, ZeroEmail). This isn't an accelerator or batch program though - no demo day, no 12-week curriculum. Just funding and hands-on help until you raise your Series A.

Why are we doing this? The warm intro system is broken. Good founders without SF networks shouldn't need to spend months working connections just to get a meeting. And you shouldn’t have to go to an accelerator just to raise capital.

We focus on vertical AI, B2B SaaS, healthcare software, infra, and devtools/open source. US-based for now.

Happy to answer questions about the process, what we look for, or early-stage fundraising in general.

"All of the under-sink reverse osmosis and two-stage filters achieved near-complete removal of the PFAS chemicals we were testing for,” Stapleton said. “In contrast, the effectiveness of activated-carbon filters used in many pitcher, countertop, refrigerator and faucet-mounted styles was inconsistent and unpredictable. The whole-house systems were also widely variable and in some cases actually increased PFAS levels in the water.” - https://nicholas.duke.edu/news/not-all-home-drinking-water-f...

It's a scanned PDF linked on that page, but someone turned it into an epub if that's what you're looking for - https://gitlab.com/sigwait/computer_connections/-/releases

This is a cool project, reminds me very much of IRC Poker from back in the early days on the internet.

We got frustrated with lack of tools available to founders to model their SAFE rounds and subsiquent dilutions, so we built an open source library and cap table worksheet.

The underlying finance library is MIT licensed and we've done our best to match it up with known sources of truth (Excel spreadsheets from lawyers), along with tests!

Feel free to give feedback to either me (mdp@1984.vc), or team@1984.vc

Related paper - https://arxiv.org/pdf/2402.03300.pdf

Paper this is based on - https://arxiv.org/pdf/2306.04529.pdf

Here’s another source - https://www.cnbc.com/2024/01/08/twilio-ceo-lawson-steps-down...

“The move is not likely to stave off the activists at Anson Funds and Legion Partners. Both are pushing for the company to sell itself or completely divest its data & applications business, CNBC previously reported.”

For those of us that didn't know the term Microtonal Music - https://www.youtube.com/watch?v=q1XOnIk2ai8

Yeah, LCWO is the gold standard for Koch training on the web, and I'd second the suggestion for CW Academy. Having a commitment to learning with other people tends to help you ride out the rough patches.

To help with my own learning, I built a mobile friendly web app/game which includes Koch as an option. It's also open source - https://github.com/mdp/morse.mdp.im - morse.mdp.im

You're right, "Sell my data" might have been too strong. But they are certainly mining it to train things like their "suggested responses". In my view, it's an ad company, and while they might not be doing it today, there's nothing stopping them from using my data in the future, hence the "free pass".

I came to a similar conclusion. You should write every email as if it were public, because it's entirely likely that it will be. They can be forwarded, made public through legal discovery, or exposed in a data breach (eg. Sony/North Korea).

Forget security for a second, imagining every email as public record will make you more considerate and less biased writer. And from a business perspective, email should be viewed as a public legal record, because in some cases it will be used that way.

That's not to say that there shouldn't be private messaging options, it's just that email isn't one of them and was never really built to be. PGP was always sort of a tacked on solution with a lot of faults (no forward secrecy, plenty of meta data leakage, usability issues)

All that being said, I still left Gmail for Fastmail. Just because I consider every email I write to be public doesn't mean I want Google getting a free pass to mine and sell my data.

This isn't much of a solution. I had a G-Suite account with them when I had an issue with Payments. I ended up with a suspended payments account all the same and no support via the G Suite support package (let's be honest, at $5 you get what you pay for). G Suite support will help you with things like problems with your email, but the Payments team is totally separate. Take a look at the support page - https://gsuite.google.com/support/ - it's pretty clear that they aren't going to help you clear things up outside of their specific area.

Same for Google Play store as a developer; didn't matter that I had a $5 account, different department so back to robo-emails you go.

That’s correct as far as I understand it. In OPs case, it seems this is what Google is accusing them of, creating a new account after other accounts were banned.

Not sure what you're using, but for anyone else looking, I've been using this - https://github.com/odeke-em/drive - to do the same

I’ll chime in for Scanbot as well. I used to use a work multi page scanner for my documents, but unless it’s more than 10 pages I just use Scanbot now. It’s got great OCR and it automatically backs up to Google Drive or Dropbox.

If you're interested in seeing how it works in action, I built a U2F demo site that shows all the nitty-gritty details of the process - https://mdp.github.io/u2fdemo/

You just need a U2F key to try it.

The LLC wouldn’t have provided much protection anyway.

https://ipbreakdown.com/blog/owner-liable-for-his-companys-c...

In general, most people misunderstand then “limited liability” of LLC’s.

I've actually found the opposite to be true. Sites seem to focus more on preventing scraping and block VPS IP's more often than commercial VPN. For example, Craigslist on Digital Oceans IP's is hilariously throttled to ~256 bytes per second. Crunchbase always tosses up a CAPTCHA. This was mostly the same for Linode.

Link the Soundcloud version, which might be easier for mobile listeners - https://soundcloud.com/evan-doorbell/sets/group-1

Hah, I also had the Alps for the same reason. I always wondered what percentage of their sales was to kids making fake IDs.

This isn't very clear when you see pages like this - https://www.amazon.com/Apple/b/ref=w_bl_hsx_s_wi_web_2528944...

Also, some Apple products are shipped and sold by Amazon, not third parties - https://www.amazon.com/gp/product/B00W5Q06VU/ref=s9_dcacsd_b...

I'm surprised Apple hasn't taken any action against them for this.

Try buying any Apple accessories. I wanted to buy the genuine Apple headphones, but it's absolutely impossible to on Amazon. "Sold by Apple" product shows up from a seller in Boston, clearly counterfeit. Report it to Amazon and they suggest mailing it back for a refund. Zero concern that the seller is labeling their goods as genuine Apple and then sending knock-offs.

Prior to this I got a set of grey market Gillette blades that were fulfilled by Amazon.

Like you, I've been with Amazon for nearly 20 years, but I've recently started being more selective about what I order from them. I'm surprised that they aren't tackling this more.