Aakar Shroff

It’s poetic to have the guy that had his tooth knocked out score the ‘Golden Goal’ to give the US their first gold medal in Hockey in 46 years.

Password managers less secure than promised | ETH Zurich

Concerning

The team conducted a study to scrutinise the security architecture of three popular password manager providers: Bitwarden, Lastpass and Dashlane. Between them, they serve around 60 million users and have a 23 per cent market share. The researchers demonstrated 12 attacks on Bitwarden, 7 on LastPass and 6 on Dashlane.

The end of sickness?!?

A single nasal spray vaccine could protect against all coughs, colds and flus, as well as bacterial lung infections, and may even ease allergies, say US researchers. The team at Stanford University have tested their “universal vaccine” in animals and still need to do human clinical trials.

Am I the only one asking during this flu season: how fast can we get this through clinical trials?

Tim O’Reilly had a conversation with Marily Nika, who started at Google as an AI PM in 2013. She discussed the tools that she uses to augment her PMing now:

One of the most exciting parts of our conversation was hearing about Marily’s rapid prototyping workflow using Perplexity for user research, custom GPTs for spec generation in her own voice, and v0 for UI mockups. With these tools, she can go from idea to functional prototype in hours, not weeks. “Every week I block time on my calendar just for AI experimentation. It’s made me a much better PM,” she said.

PMs and non-PMs should get comfortable with this type of workflow now because this is going to be how our jobs are going to evolve.

This post on what it’s like to work at OpenAI is fascinating. A wealth of great insight into the craziness of one of the world’s fastest-growing companies. One thing that stood out to me was this about CODEX:

From start (the first lines of code written) to finish, the whole product was built in just 7 weeks.

Now, to be fair, they’d been thinking about it since November 2024, but still just “7 weeks”! I’ve been using CODEX for the last few weeks instead of Cursor or Claude Code to tinker around, and it’s really fun to just have it go off and do work.

An interesting look at the way Medium got back to profitability. Being able to get a first-hand view of the turnaround provides great insights:

We owed those investors $37M for loans that were overdue. Folks, on paper that means we were insolvent. The investors also held an additional $225M of liquidation preferences.

I hadn’t heard about the idea of recapping, so that was new.

Christina Caron in the New York Times elaborates on a new study on what makes 🆒 … 🆒

A new study suggests that there are six specific traits that these people tend to have in common: Cool people are largely perceived to be extroverted, hedonistic, powerful, adventurous, open and autonomous.

Well then

These researchers though are really standing out here on a limb:

Coolness that involves risk-taking and being socially precocious during adolescence may offer popularity during youth, but one study published in 2014 found that many teenagers who behaved in this way would later struggle in their 20s, developing problems with alcohol, drugs and relationships. “They are doing more extreme things to try to act cool,” one of the researchers told The New York Times.

Christopher Mims recently wrote about how Clorox and its employees are using generative AI to power “new R&D ideas” and “better customer insights.” These tools are only going to get better at automating away repetitive work—but for now, I think the real value is in using them to augment your day-to-day. You still need to be in the loop.

While AI does pose a threat to workers if it’s used by companies to automate away jobs, usage of it in corporate America is still largely grassroots: Employees are grabbing available tools to augment their capabilities and enhance their work, say workplace researchers.

And that’s necessary, because the tools—while improving seemingly by the hour—still lack the nuance to know when something’s just… wrong 💩:

“If you go in with the expectation that the AI is as smart or smarter than humans, you’re quickly disappointed by the reality,” says Eric Schwartz, Clorox’s chief marketing officer.

During brainstorms, the AI tried to push the idea of “bleachless bleach,” he adds, which isn’t something that would actually work in real life. It’s a nonstarter, especially at Clorox.

When brainstorming about cat litter, the AI pushed the idea that since you love your pet, you might also love your pet’s poop. It takes a human to realize, “No, that wouldn’t sound good,” he says.

But in the more repeatable parts of the workflow, generative AI really shines. Like review analysis:

Clorox uses an AI-powered analytics tool that scans reviews and ratings of its products from Amazon, Walmart and dozens of smaller retailers. The company’s systems can summarize those reviews and dive into “attribute based sentiment analysis,” to learn what aspects of a product really resonate.

All of that’s interesting—but the stat that floored me comes right at the top of Mims’ article:

Hidden Valley Ranch needed a new formula. No, the recipe for America’s favorite condiment wasn’t changing. After all, last year it beat ketchup in sales.

Ranch > Ketchup. Who knew?

I didn’t plan to spend the weekend with “The Bear”, but here we are.

After a shaky and somewhat meandering Season 3, I was honestly on the fence about diving into Season 4. I kept asking myself if it was even worth it.

But curiosity won out. I hit play on Friday… and didn’t stop until it was over.

And I’m glad I didn’t.

Season 4 pulled me back in. It was still emotionally intense — the kind of anxiety-inducing storytelling that The Bear has become known for — but it felt more grounded, more purposeful. The season had a real arc. It started strong, ended stronger, and managed to recapture some of the spark that made the earlier seasons so compelling.

Here’s hoping we don’t have to wait too long for Season 5.

If you own a Kindle you’re probably like me, you send personal documents such as PDFs and books purchased elsewhere to your Kindle library via email so you can read them on your other devices. I primarily use my Kindle Paperwhite, so most of my highlights end up in the My Clippings.txt file stored locally on the device. That file is easy to export: just plug in the Kindle, grab the text file, and email it to Readwise to sync all my new highlights. Simple.

But recently, I’ve been reading across multiple devices. For example, when I forgot my Kindle at home, I caught up on a doc using the Kindle app on my iPhone. Since I had emailed the document to my Kindle address, it appeared in my library and downloaded smoothly on the app.

To my surprise, the highlights I made on my iPhone synced perfectly to my Kindle Paperwhite. That part worked beautifully. But when I went to export my highlights to Readwise, there was a problem: those synced highlights weren’t included in My Clippings.txt. Since I didn’t make those highlights on the Paperwhite, they never made it into the export file.

Amazon offers a way to email highlights from the Kindle app, even for personal documents but there’s a limit to how many highlights you can send. And honestly, restricting myself to only highlighting on one device just to get around this limitation felt wrong. So I decided to dig deeper.

If highlights made on the iPhone are syncing to the Paperwhite, then they’re being stored locally somehow. The question was: where?

Fortunately, the community at MobileRead forums has done a great job reverse-engineering Kindle file structures. In particular, jhowell has created two essential tools:

• KindleUnpack: Reads metadata from Kindle file formats
• KFX Input Plugin for Calibre: Allows decoding of .kfx files used in newer Kindle documents

Using those tools—and with some help from ChatGPT Codex—I built a Python script that can extract synced highlights from the Kindle’s local files. You’ll need two files for each personal document:

1. The .kfx file (the main book or article)
2. The corresponding .yjr file (which stores your annotations)

Once you have those, the script will parse the highlights and generate an HTML file you can review or import into another tool like Readwise.

I’ve posted everything here:

👉 GitHub Repo – KFX Highlights Extractor

I’ve tested this on a few documents, and it worked well, but your mileage may vary. If you run into issues or have improvements, feel free to open a pull request or file an issue.

This story regarding the Graphing Calculator app that Apple bundled into the first PowerPC computers is wild! Imagine being so overtly obsessed and passionate about something that you figure out how to keep working on a cancelled project without pay and by continuously sneaking into the office that “fired” you.

🤯

Happy WWDC week to all the Apple developers!

The Atlantic’s Olga Khazan points out that our minds have a remarkable knack for letting the pain of past sorrows fade faster than the warmth of joy.

people tend to romanticize the past, remembering it more rosily than it actually was. Thanks to something called the “fading affect bias,” negative feelings about an event evaporate much more quickly than positive ones.

Just remember to be careful telling people who are in “it” that time heals all wounds quickly.

Thomas Ptacek with his post ”My AI Skeptic Friends Are All Nuts” is making the rounds and I have to say it’s great to see. Everything he writes encapsulates how I feel about the current times. From the responsibility:

Are you a vibe coding Youtuber? Can you not read code? If so: astute point. Otherwise: what the fuck is wrong with you? You’ve always been responsible for what you merge to main. You were five years go. And you are tomorrow, whether or not you use an LLM.

To the job:

Reading other people’s code is part of the job. If you can’t metabolize the boring, repetitive code an LLM generates: skills issue! How are you handling the chaos human developers turn out on a deadline?

To the future:

So does open source. We used to pay good money for databases. We’re a field premised on automating other people’s jobs away. “Productivity gains,” say the economists. You get what that means, right? Fewer people doing the same stuff. Talked to a travel agent lately? Or a floor broker? Or a record store clerk? Or a darkroom tech?

The entire post is a worthwhile read so do yourself a favor and make some time to read it.

Ethan Mollick recently posted about how he’s got the o3 model to spit out New Yorker like cartoons. I thought I’d give it a shot for something that I need to have humor over:

Create an image make me a funny, wry, and original New Yorker cartoon about the Celtics being down 2-0 in the playoffs against the Knicks. Start with 20 ideas and pick the funniest and most New Yorker like and make that.

Morningbrew had a story about a man who purposefully allowed snakes to bite him:

Tim Friede first allowed a venomous cobra to bite him in 2001 with the aim of building up his own immunity as he pursued his hobby. Over time, his goal broadened. He wanted to raise awareness for snakebite treatment and prevention

How many people would walk the walk:

Friede experienced 200 bites from “all manner of venomous snakes” and injected himself with the venom of 700 specimens, including some of the world’s deadliest, he said.

As a result of his generosity we now have significant improvements for anti-venom:

Thanks to said blood, Centivax created an antivenom that research now shows offers “unparalleled” protection against 13 lethal snakes and partial protection from six others.

I remember watching this pizza throw on replay so often.

The only thing more Boston would have been if the throwers name was Sully.

A few weeks ago I posted about a New York Times article highlighting Rippling suing Deel, well get your popcorn ready because the affidavit is out. And let me tell you it reads like a script that Netflix has already optioned. There’s a bunch of juicy details in it:

“Alex told me he ‘had an idea.’ He suggested that I remain at Rippling and become a ‘spy’ for Deel, and I recall him specifically mentioning James Bond.”

Never good when the corporate espionage plan starts with a 007 reference on a WhatsApp call from the office conference room.

“I asked Alex and Philippe to paper our arrangement with a consulting agreement… Alex refused… I understood that his refusal was due to the fact that he did not want a documentary record of our relationship.”

Airtight strategy: no paper trail, just vibes.

“Philippe established payment terminology… I would send a picture of a watch… Philippe would say ‘send that watch to London.’ Then he would say ‘the buyer is happy.’”

Corporate code language so weird it sounds like a Montblanc commercial directed by Wes Anderson.

“Among the specific searches that I recall Alex asked me to run were searches for ‘tom brady,’ ‘iran,’ ‘sanctioned countries,’ and ‘tinybird.’”

Tom Brady and sanctioned countries. Normal competitive intel stuff. As a Patriots fan this is the weirdest search here. Why Tom Brady? Is there some celebrity endorsement deal? Is Tom Brady looking for a global HR solution?

“I smashed my old phone with an axe and put it down the drain at my mother-in-law’s house, as Deel’s lawyer Asif had advised.”

There’s a lot to unpack in this sentence, but the axe is really doing most of the work.

“Deel’s lawyer Asif said… ‘don’t worry, PSG is going to sort all this out.’ I understood him to be referring to Alex.”

PSG: not Paris Saint-Germain, but apparently a new alias for Deel’s CEO. Although Deel’s CEO really likes PSG, Paris Saint-Germaine, so who knows?

“I decided to cooperate after I got a text from a friend on March 25, 2025 saying, ‘the truth will set you free.’”

Spy thriller meets after-school special. When does this appear on Netflix? Just want to know when to get my 🍿 ready.

I really recommend reading through all 12 pages of the affidavit you won’t be disappointed.

Fun times here in NYC the next few days. ☔️🌧️

Rippling Sues Deel, a Software Rival, Over Corporate Spying - The New York Times

Rippling said it obtained a court order last week forcing D.S. to turn over his phone. But when a court-appointed lawyer showed up at Rippling’s Dublin office and demanded that the employee hand over the device, D.S. locked himself in a bathroom. He later fled the scene, it said.

Hard to say you aren’t guilty after doing that.

Molly White pointing out that free and open access is still a necessity even in a generative AI world. And for everyone to be successful AI companies need to figure out a structure that will allow them to give back:

The future of free and open access isn’t about saying “wait, not like that” — it’s about saying “yes, like that, but under fair terms”. With fair compensation for infrastructure costs. With attribution and avenues by which new people can discover and give back to the underlying commons. With deep respect for the communities that make the commons — and the tools that build off them — possible. Only then can we truly build that world where every single human being can freely share in the sum of all knowledge.

My favorite part though was this nice jab:

Many AI “visionaries” seem perfectly content to promise that artificial superintelligence is just around the corner, but claim that attribution is somehow a permanently unsolvable problem.

My Whoop telling me that I might be living too close to the edge.

This is a case that “reply all” is funny for us all.

“Dear Boomers,” Tina Fey wrote in the thread at 9:38, with a link to an instructional YouTube video for senior citizens on how to read, reply, and forward emails when using a Gmail account.

Electric first 9 seconds of the USA 🇺🇸vs. Canada 🇨🇦 hockey game from last night. I can’t state this enough: this is an all-star game! Good win by the 🇺🇸.

My eyes glazed over towards the end of this post on the complexity of writing a calculator app. But it’s really fascinating. Hans-J. Boehm and his team developed an innovative approach using a combination of rational arithmetic and recursive real arithmetic to handle both exact and irrational numbers efficiently.

To give correct answers to mathematical expressions, a calculator must represent numbers. And almost all numbers cannot be expressed in IEEE floating points.

I’ve been a subscriber to Flighty for a few years now. It’s not only a well done app from a visual design perspective but it’s also an insanely fast provider of flight data. In a lot of cases before your airline or the airport provides it to you. As an example, I was traveling back to the states from Mexico, the airport I was flying out of doesn’t post the gate number of your flight until one hour before takeoff. However, Flighty was able to tell me the gate 10 minutes before United notified me via the app and via SMS. I have no idea where and how it’s pulling the data but it’s great!

I’m often flying once a month for work so the subscription is a no-brainer.