I’m back from my travels into a productive week at work.

I still have a lot in my “to read” pile, but today, I have an array of topics to share. So buckle up while we delve into this week’s newsletter.

What’s Happening?

Yes, I’m human, and I just used “delve.” What’s the big deal?

Is AI Worth It?

Well, according to Y Combinator co-founder and essayist Paul Graham, using “delve” suggests that the text is likely written by AI.

Specifically, PG shared a graph showing a massive spike in the frequency of the word in research articles since the proliferation of large language models and AI.

Surprisingly, this Tweet received a lot of backlash from the Nigerian community. PG put out these fires, clarifying his preference for simpler words that convey exact ideas over more complex and unnecessary ones.

When the smoke cleared, it exposed the dark underbelly of the AI race, built on cheap labor from English-speaking countries like Nigeria and possibly other African and Indian subcontinent countries.

Someone from these countries likely went through a lot of text and images, sometimes traumatizingly graphic, and classified them by hand to generate the training data for the models that power the AI models we use to prompt cute pictures of puppies.

Unsurprisingly, the companies that build and benefit from these models keep their pockets full and their hands clean.

Molly White’s article “AI isn't useless. But is it worth it?” illuminates this discussion from a different perspective, in which AI unfailingly fails to live up to the hype.

This isn’t new, however, and is expected with all technological triumphs, like the blockchain/Bitcoin hype, which revealed itself to be a pump-and-dump scheme.

Similarly, Devin made tech influencers preach the end of programmers. For a week. It was recently revealed that their whole demo was fake.

This gets even better. Devin is now being outperformed by an open source alternative from the University of Singapore.

Most VCs don’t understand open source, but they have to be enlightened real quick because open source is eating software.

Chad Zuck

Speaking of open source and AI, Mark Zuckerberg dropped a banger on Dwarkesh’s podcast, where he talked about Meta’s open source Llama models.

While open source business models have been around for decades, he rightfully points out that open source for businesses is still novel and hard to comprehend for VCs.

Zuck trailblazing the open source AI movement while looking like an absolute chad sparked some fiery tweets.

Dwarkesh is an excellent interviewer and podcast host. I started listening to podcasts during the COVID lockdowns, and I often learn a lot from interesting people having interesting conversations. Dwarkesh takes it to another level, and the effort he puts into each interview is evident in the outcome.

And he is 23.

Company-Killing Reviews

Marques Brownlee’s blatantly honest review of the Humane Ai Pin has been the talk of the town for the last few weeks.

The video’s title, “The Worst Product I’ve Ever Reviewed… For Now,” sums up his opinions on the product, which has been in the spotlight for its novelty in the form factor ever since it was initially announced.

There are enough people on the internet critiquing Marques for his critique, claiming that it could kill the company. This prompted Marques to release a new video asking the question, “Do bad reviews kill companies?”

Well, they don’t. But bad products do. And apparently, this is difficult to understand.

There was a similar uproar by producers in the Malayalam Film Industry recently against movie reviewers. Their stupid claim was that their movies fail because people give bad reviews. If only they made better movies!

This claim also insults Malayalees, renowned for their taste in films across India.

People pay money to buy your products and spend time using them. Whether it is a movie or an AI pin, people deserve to know what they are getting into. I repeat: bad reviews don’t kill companies; bad products do.

I have a lot more to share, but you should definitely check these out:

  • The Domino Computer: An excellent interactive article walking through Matt Parker’s Domino Computer demonstrating how simple operations can be used to build extremely complex logic in modern computers.
  • The Forgotten War on Beepers: Pagers were linked to drug trade? This article explores the forgotten legislation against pagers in schools to “save” the youth against their questionable usage. This closely mimics the narrative around smartphones in schools.
  • The Extended Internet Universe: Venkatesh Rao explores the different layers of the internet and how we are increasingly retreating to the “cozy” comfortable parts of the internet as the broader internet turns into turmoil.
  • I Made a Graph of Wikipedia... This Is What I Found: This video explores the connections between Wikipedia articles and what these connections reveal. What it fundamentally reveals is the nature of the broader internet.
  • 4.5 million times faster than average broadband: In optimistic technology news, we might have much faster internet pretty soon. What could this mean for the internet and content as we know it when the limits of the current network are no longer a problem? I’m pleased about not having to worry about my PageSpeed Insights.
  • WebAssembly Can't Win: I have already been labeled as a Golang absolutist, and I might also be labeled a WebAssembly absolutist. While I primarily run Wasm on the server, I have also been exploring Wasm in the browser for some of my personal projects. But this article throws all that out of the window and says that Wasm is only useful on the browser if the binaries are small.

Hot off the Press

I wrote a daily log entry from an airplane last week.

I was on my way back from Hanoi after the FOSSASIA Summit, and I took some time to reflect on the conference and my interactions with the people and the country. It also has some photos.

Read here: “#279 Airplane Logs (FOSSASIA Summit 2024)”