Every week, get the latest curated high signals on news, insights, and ideas around technology, science, and business to help you become a better builder and thinker. Please fill out the survey so we can deliver the best content to your inbox. Thanks for reading!Thanks for reading!
Did you know
Fred Smith saved FedEx by gambling the company’s last $5,000 in Las Vegas and winning $27,000 at blackjack.
Fred Smith, the founder of FedEx, is famous for a gamble that helped save his company in its early days. In the mid-1970s, FedEx was on the brink of bankruptcy, with only $5,000 left in its bank account. After being denied a crucial business loan, Fred took the company’s last funds to Las Vegas and played blackjack. He actually won $27,000 which was enough to pay a vital fuel bill and keep FedEx operating for another week. This lifeline allowed Smith to secure an additional $11 million in investment, which ultimately helped FedEx survive and grow into a global logistics giant. Smith later reflected:
“No business school graduate would recommend gambling as a financial strategy, but sometimes it pays to be a little crazy early in your career.” 😲
This story has become a legendary of founder risk, but probably not the best advice or journey to emulate for most. Nontheless, it’s such a bad-ass and memorable story showing what extremes risks founders are willing to take on their entrepreneurial journey.
Here’s this week’s scoop:
Using AI to Fight Cancer
China’s Beating the Nuclear Cost Game (and What Everyone Else Can Learn)
A football club is now using AI to sign players faster
Reddit is trying to become a search engine — finally!
Aeneas: How Google DeepMind is using AI to decode the ancient inscriptions
Personal Superintelligence — Essay by Zuck (Meta) that focuses on directing AI towards individual empowerment and a super intelligent assistant in their pocket [LINK]
Google’s Web Guide — An experimental AI-organized search results page that dynamically presents information based on your search query [LINK]
How Netflix Streams Live Events — A technical blog that takes you through Netflix’s journey into the engineering architecture behind their live events feature. [LINK]
Frontier Technologies Trends Outlook for 2025 (McKinsey) — an in-depth report on 13 frontier technology trends [LINK]
Dark Patterns: tricks to make you spend more online — a guide to help you better spot dark patterns and avoid them when shopping online [LINK]
📡 The Signal
Using AI to Fight Cancer
Cancer is notoriously hard to beat because our immune system’s T-cells (a type of white blood cell that fights against infections) can’t find or recognize cancer cells very well. Normally, scientists try to boost these T-cells by searching for natural proteins in our cells that can help the immune system spot cancer. But usually, the process is slow, takes months, and sometimes doesn’t even work at all. Recently, new research has shown a new way to solve this problem using AI. So instead of hunting for natural proteins, scientists used AI to design brand new proteins from scratch that acted like a GPS for T-cells, helping them find and attack cancer cells directly. They trained the AI models on known protein structures and then had the AI suggest new protein shapes and sequences that could fit cancer targets. They found that T-cells equipped with the AI-designed protein killed melanoma cells much faster and stopped the cancer from growing!
The cool animation below shows red melanoma cells getting destroyed by immune cells with these new proteins, which really shows their impact. This is a big deal because we’re now seeing genuine use cases of AI being able to help make new medicines faster and more effective. The fact that the process is so much faster and more flexible could mean better and more personalized cancer treatments down the road. 🚀 [LINK]
China’s Beating the Nuclear Cost Game (and What Everyone Else Can Learn)
Nuclear is making a BIG comeback, and everyone’s suddenly back on this bandwagon. But there’s a huge problem where building nuclear plants usually gets more expensive the more you build, which is the exact opposite of what happens with solar and wind. The way it works is that normally, you try to fix this by making bigger plants or learning from experience, but that hasn’t really worked in places like the US or France. In fact, in these countries, costs keep going up, especially after big nuclear accidents. You can see in the chart below how prices in the US/France shot up over time, while China’s prices stayed pretty low.
But what’s different about China? It turns out that instead of just copying what others did, they actually focused hard on building up their own supply chains and keeping their regulations stable. They started by bringing in foreign tech and making the simple stuff themselves, then gradually moved on to producing more complicated parts at home. This meant they didn’t have to buy expensive imports and make stuff for way less money, where even some parts are literally half the price compared to buying them abroad. The other part to this is that China also made sure its rules didn’t keep changing and gave steady support to the companies building the plants, which helped keep costs predictable. The bottom line is that building more nuclear plants isn’t enough; you also need a solid plan for making your own parts locally and keeping rules steady. It just goes to show that China’s success isn’t just really about the tech, but more about sticking to a plan, building up local skills, and not messing with the rules every few years. The nuclear race is going to be interesting, but it’s clear China is in the lead. [NATURE]
A football club is now using AI to sign players faster
Managing player contracts in football is usually a huge pain. Clubs spend ages going back/forth with lawyers, reviewing the same documents over and over, and waiting weeks just to get things signed. This slows the transfer and keeps staff tied up in paperwork, which could result in players being signed on by other clubs that move much faster. Recently, football club Cambridge United partnered with Genie AI to start using AI to handle all its contracts. The workflow drafts, reviews, and tracks contracts automatically. It’s the first pro club to run all player contacts through AI. Tests show that what used to take weeks now takes days, which is a massive improvement with transfer windows closing soon. We’re going to see more of this in other sports, i.e., NFL, NBA, Baseball, Cycling, etc. It’s just a matter of time. [LINK]
Reddit is trying to become a search engine — finally!
Reddit is quickly becoming the go-to platform for real, authentic, and genuine content. When people want real answers from real people, they usually just Google their question and tack on “Reddit” at the end. This is important because Google’s search results are often full of ads, random websites, and now AI-generated content that doesn’t feel helpful or human-like. It’s now getting worse because Google’s getting even more into giving AI answers, which means it might send less traffic to Reddit in the future. Normally, if you own a website, you’re just hoping Google keeps sending you visitors, but that’s risky because if Google changes its mind, you’re out of luck.
Reddit is now considering beefing up its own search tools so people can find what they need right on Reddit. It looks like they already have over 70 million people using Reddit’s search every week, and their new AI-powered tool, Reddit Answers, is catching on quickly, jumping from 1 to 6 million weekly users in just a few months! In a world full of AI-generated content, Reddit is probably one of the last standing platforms that still continues to generate authentic content. I think it’s a smart move for them and not to be beholden to Google’s algorithm. If they pull this off, it could change how we find/search for advice online. [LINK]
Aeneas: How Google DeepMind is using AI to decode the ancient inscriptions
Animation of a restored bronze military diploma from Sardinia 113/14 C.E. (CIL XVI, 60).
It’s a huge challenge piecing together history because discovered artifacts like old texts have aged poorly, with missing pieces, faded words, and even those that have been scratched out. Usually, experts have to spend a long time matching up these fragments with similar texts, trying to fill in the blanks or figure out where and when they were written. It’s slow and frustrating work, with even the best historians struggling.
So instead of slogging through piles of inscriptions, a new AI tool from Google DeepMind called Aeneas can scan thousands of them in seconds that would take humans much longer to find. It doesn’t just look at words, but rather, it can use pictures of the inscriptions to help figure out where they came from. It’s also the first tool that can guess missing chunks of text even when we don’t know how big the gap is, which is a big deal for really damaged artifacts. They put Aeneas to the test on a famous inscription about Emperor Augustus, and the results matched what historians have argued about for years, but Aeneas did it way faster and showed why it made the choices it did. This could hopefully shed some new light on historical pieces that were once lost to history, and have a new meaning, which may provide more context for that time. [DEEPMIND]
Here’s a direct quote on how Aeneas works:
Aeneas is a multimodal generative neural network that takes an inscription’s text and image as input. To train Aeneas, we curated a large and reliable dataset, drawing from decades of work by historians to create digital collections, especially the Epigraphic Database Roma (EDR), Epigraphic Database Heidelberg (EDH) and Epigraphic Database Clauss Slaby (EDCS-ELT).
We cleaned, harmonized and linked these records into a single machine-actionable dataset that we refer to as the Latin Epigraphic Dataset (LED), comprising over 176,000 Latin inscriptions from across the ancient Roman world.
Our model uses a transformer-based decoder to process the textual input of an inscription. Specialized networks handle character restoration and dating using text, while geographical attribution also uses images of the inscriptions as input. The decoder retrieves similar inscriptions from the LED, ranked by relevance.
Intel is still on struggle street
Intel is going through some major cost-cutting, especially in its chip-manufacturing business, which is basically its huge bet to get back in the game. Usually, when a tech company hits a rough patch, it’ll try to spend its way out by investing in new projects and hoping something will stick. But it looks like this hasn’t really worked out for Intel. They spent a ton on new factories + tech, but didn’t really have enough customers lined up, so now they’re stuck with half-empty facilities and wasted money. So now, their new CEO (Lip-Bu Tan) is reigning in the spending and only building what customers actually commit to buying. Apparently, if they can’t get a big customer for their next-gen chip process (called 14A; their 1.4nm class), they might have to give up on the foundry business altogether. That’s a huge deal because it shows Intel is willing to walk away from a business that’s not working, instead of hoping for a miracle. Their net loss is getting bigger, so it looks like extreme measures have been put in place by cutting projects and 15% of the workforce. The silver lining is that Intel has been brutally honest about its struggles, which we don’t see often. The CEO’s new approach is: let’s cut our losses and stop pretending everything’s fine. [LINK]
Humanoids are going exponential
Making robots that can actually work with humans, do real jobs, and not be creepy has been a big challenge. Usually, robots are either too expensive, weird, or just not useful enough for everyday life or work. With recent advancements in GenAI, there’s a new emerging arms race in robotics. Tesla (Optimus) and startups like Figure and Unitree (from China) are actually putting humanoid robots into the real world. The innovation and progress in this field have heated up not just with the amount of money put into this sector, but also how much the improvements that have been made to the movement and dexterity of these robots. Companies are even hiring people to teach robots to move more naturally, and researchers are focused on making robots seem friendly. Watching the Unitree R1 robot demo almost feels like it was done by CGI or AI-generated, but it’s crazy to see how agile these robots have become. Just last year, robots could barely run; now they’re running and doing martial arts. [UNITREE, SCMP]
Figure 1 Robot
If you liked this, here are a few more popular posts:
To build the best newsletter for you all, I’d love to get some feedback. If you have some time, please fill out the following survey.
🏆 Alpha Bytes Premium
If you like our content, you can explore our Premium option. You can read more about it here. This includes (for now):
✅ Full access to the weekly newsletter ✅ Full access to podcasts + curated transcripts in Q&A format—this has become widely popular. ✅ Full access to all deep dives, essays, and tactical field guides ✅ Early access to new features and experimental content
🤩 Deals, Discounts & Affiliates
You can find carefully selected offers and deals here.
