Hey there Bizarro readers!
For those of you who are new here - welcome. For our returning and long-time readers - it’s nice to see you again. Grab your favorite beverage, get comfortable, and enjoy this month’s lineup of stories, stats, tools, and more. Let’s kick it off with our headliners:
⚡🎬 5-Second AI Video = 1 Hour Microwave Time
🤖📝 Students Are Now Asking AI to Add Typos to Fool Teachers
🖨️🫀 Scientists 3D Print Inside Bodies Without Surgery
Enjoy and say hello in the comments. Don’t worry, no one’s going to bite you. Maybe.
MIT Technology Review recently conducted a comprehensive new investigation into AI's energy appetite and their analysis revealed that the technology is quietly reshaping America's entire power grid - and we're largely in the dark about the true scope of what's coming.
Energy costs of using AI vary wildly depending on what you're asking it to do. Simple text responses range from running a microwave for 1/10th of a second to 8 full seconds. In comparison, generating a high-quality image takes ~5.5 seconds of microwave time. However, all of those pale in comparison to creating a 5-second video, which requires over an hour of microwave energy!
By 2028, AI alone could consume enough electricity to power 22% of all U.S. households if that power were redirected to homes instead. Meanwhile, data centers overall (which include AI plus everything else) are projected to grow from their current 4.4% of total U.S. electricity consumption to 12% by 2028.
U.S. citizens might end up indirectly subsidizing this AI revolution even if they don’t use a single AI tool. New research shows that utility deals with tech companies often pass costs to regular consumers through higher electricity bills. For example, Virginia residents could see an extra $37.50 monthly charge to subsidize new data centers.
The most troubling aspect of this is that we're flying blind into an energy-intensive future because AI companies won't share real usage data, claiming trade secrets. As one researcher put it: "All bets are off in the coming years" as AI agents, reasoning models, and personalized AI multiply our current usage exponentially. What started as innocent chatbot queries is rapidly becoming one of the biggest energy stories of our time.
The academic cheating arms race has reached new levels of absurdity. According to a New York Magazine investigation, students aren't just using AI to write their papers - they're coaching chatbots to write poorly on purpose, complete with typos and freshman-level mistakes, to avoid detection by teachers and AI-detection software.
The real overachievers have developed a multi-step "laundering" process: write an essay with one AI tool, feed that output into a second AI chatbot, then run it through a third one to dilute any telltale AI patterns. It's like money laundering, but for intelligence.
The strategy doesn't always work, especially when students know so little about their subject that they can't spot AI nonsense. University of Iowa TA Sam Williams watched his class turn in heartfelt personal essays one week, then submit AI-generated reports about "the history of New Orleans jazz featuring Elvis Presley" the next week. Elvis, of course, had nothing to do with jazz or New Orleans.
Some teachers are fighting back by going old school - bringing back handwritten Blue Book exams where students must write essays in class with pen and paper, eliminating any chance to run their work through multiple AI tools.
While I don’t condone cheating and I think that ultimately these students are doing themselves a disservice, there is also a counterpoint to consider here. On the one hand, those who cheat using AI tools are missing out on the knowledge they would have gained otherwise. Not to mention they are paying college tuition to not learn. But on the other hand, they are training themselves to become efficient prompt engineers. Maybe there’s a happy middle ground? What do you think?
|
| Do you believe it's possible for educators to find a way for students to still continue to learn while also using AI in a responsible way? | |
|
|
Forget everything you know about medical implants and tissue repair. A team from Caltech just unveiled technology that can 3D print tissues, drug depots, and biosensors directly inside your body using nothing more than an injection and ultrasound waves. This breakthrough could eliminate the need for invasive surgeries to place implants or repair damaged tissues.
The system, called DISP (deep tissue in vivo sound printing), uses a special "sono-ink" that stays liquid at body temperature but solidifies into precise structures when blasted with ultrasound. Unlike light-based 3D printing that can barely penetrate skin, ultrasound can reach nearly 8 inches deep into organs without causing damage.
The researchers used DISP to create drug depots that slowly release cancer medications directly at tumor sites (instead of being flushed away in hours like current treatments). They also printed electronic biosensors to monitor tissue activity, and even repaired tissue damage.
The injectable ink includes multiple smart components: molecules that link together when triggered, fatty bubbles filled with binding agents that release when hit with ultrasound, and tracking molecules that light up so doctors can monitor the printing process. Any excess ink is safely broken down by the body's natural processes, and the technology doesn't trigger immune responses.
While there are still challenges (like printing on moving organs such as lungs and hearts), researchers believe AI could help solve these issues by rapidly adjusting to changes in the body during printing. This isn't science fiction anymore - it's the future of personalized medicine, where your body becomes its own 3D printer for exactly the treatments you need, exactly where you need them.
Sponsored by Optimole, the best image optimization tool on the internet.
Researchers have developed a new laser amplifier that can transmit information 10 times faster than current fiber-optic systems by expanding the bandwidth from 30 nanometers to 300 nanometers. The silicon nitride amplifier uses spiral waveguides and could improve internet speeds while enabling more precise medical imaging and diagnostics.
Modern web frameworks like Next.js can use React Server Components to build completely static websites by running server code during the build process and saving the outputs as static files. This "hybrid" approach gives developers the benefits of server-side features while allowing for free static hosting.
A senior React developer shared their experience with a technical interview at a major tech company, detailing a 60-minute assessment where they built a "Dress Sales Tracker" application with progressively complex requirements.
Stack Overflow, once the go-to hub for programmers seeking help, has hit a grim milestone. Recent data shows question volume has plummeted to levels not seen since the site's 2009 launch. This dramatic cliff dive signals what might be the final chapter for a platform that revolutionized developer collaboration.
A lot of organizations are using AI at work these days, but far less have explained exactly how they’re using it. Mark Zahra wrote a useful blog post on this topic, where he peeled back the curtains on how WP Mayor has been leveraging AI to improve their workflows. Check it out for ideas you can borrow.
Klarna CEO Sebastian Siemiatkowski admitted the company's AI-first customer service strategy that replaced 700 human workers led to lower quality support, prompting a strategic pivot to rehire human agents through a remote, on-demand model.
Scientists have developed a laser system that can read text in a book from 0.85 miles / 1.37 km away by using intensity interferometry - a technique normally used in astronomy to measure distant stars - with eight infrared laser beams that can detect individual letters just 3 millimeters across.
🎥🎞️ At a recent E.U. summit, the host country of Albania, welcomed the visiting leaders of European countries with an AI-generated video depicting all of them as toddlers. It seemed to be a hit among the crowd and the leaders themselves, but definitely qualifies as bizzaro.
The annual I/O developers conference hosted by Google resulted in over 100 new AI tools and features being released (or ones that will soon be released). I tested a few of them. Here are my top six, as well as my thoughts on Google’s AI strategy.
Have you ever wanted to run multiple domains off of a single WordPress installation? If you didn’t think that was even possible - it actually is. All you need is the WP Landing Kit plugin. Check out the breakdown of how it all works.
“Nobody expects a computer simulation of a hurricane to generate real wind and real rain. In the same way, a computer model of the brain may only ever simulate consciousness, but never give rise to it.”
– Anil Seth, a neuroscientist, author, and public speaker who has pioneered research into the brain basis of consciousness for more than 20 years.
63% of people think that Google search results were better the previous year, according to a recent WalletHub study published in February 2025 that evaluated search results for financial products. The study found that only 41% of Google's top 10 search results actually meet user intent, and trusting Google's top 5 results could cost consumers an average of $202 due to suboptimal financial product recommendations.
73.6% of Alphabet's (Google’s parent company) $237.9 billion in ad revenue came from Google Search in 2023 - totaling $175.0 billion according to Yahoo Finance.
200,000,000 years old is how old the world's most heavily guarded tree species is. The Wollemi pine, a "living fossil" was rediscovered in 1994 in Australia, but there are fewer than 100 mature trees left in the wild. They are classified as critically endangered and remain under strict biosecurity protection against pathogens and poaching.
£40,000 - £60,000 is the estimated range that Alan Turing's rare signed 1939 PhD dissertation is expected to fetch at auction, as part of a remarkable archive of the computing pioneer's papers that was nearly destroyed during an attic clear-out before being rescued by family members in 2024. The entire archive goes to auction this month.
Motion: This open source motion library boasts first-class APIs for both JavaScript and React. It combines the power of JavaScript animations with the performance of native browser APIs.
https://motion.dev/
React ChronoSmart: This React timeline component offers versatile display options with three viewing modes (Horizontal, Vertical, and Tree). It features a straightforward data-driven API, fully customizable styling, TypeScript support for better code quality, nested timeline capabilities for complex hierarchies, and built-in slideshow functionality with various animation effects. Perfect for creating visually appealing, interactive timelines in your React projects.
https://react-chrono.prabhumurthy.com/
Pages CMS: This GitHub-based content management system makes updating static sites painless without wrestling with Git or YAML files. Perfect for Next.js, Astro, Hugo, and Nuxt projects, it provides a friendly interface with visual editing, media management, search capabilities, and content scheduling - all running directly on GitHub. It's completely free to use and works with many popular static site generators.
https://pagescms.org/
Cliché, but true: perspective is everything. Next time you feel pressure, remind yourself that in an alternate reality, you could be this guy.
積ん読 (Tsundoku) is a Japanese word that translates into the act of buying books and letting them pile up unread. Something akin to a book hoarder.
We recently turned on the pledges ask in Substack. Here’s why:
Bizarro Devs has been a free publication for 168 issues, but none of those issues have been free to produce. As a company, we have absorbed the cost because we wanted to give back to the developer community that we are also a part of.
Unfortunately, Google’s algorithm changes in the past year have put a significant dent in our revenue, which has made it more challenging to continue operating “as is.”
There is a very real possibility that we will no longer be able to sustain the publication of Bizarro Devs on our own past the summer. We turned on the pledges to see if our community here would be willing to help us keep the newsletter alive.
Regardless of the response, we don’t plan on immediately converting the newsletter into a paid publication, but the next two or three months will determine how we chart our course and the pledges will play a role in that.
Thanks for considering, and as always, thank you reading!
If this is the first time you're here at Bizarro Devs and you enjoyed the content, then sign up here to join us on the first Tuesday of every month.
Until we see each other again,
– Martin D.
Tidak ada komentar:
Posting Komentar