A newsletter by developers, for developers, curated by Cassidy Williams and the Stack Overflow team. Every week, we’ll share a collection of great questions from our community, news and articles from our blog, and awesome links from around the web.
This week, we're thinking about the unsimple things in life. For instance, what are we supposed to do with a "dream machine" when its dreaming turns into hallucinations? How are 256 electrodes placed on the brain able to create perfect speech? Can you replace several cables with a single fiber line and still have internet? Are Ghibli breads real, and if so, where can I get one? You know, just everyday ponderings. XX Over on the pod, Stack Overflow legend Jon Skeet joined us to impart some of the tech wisdom that garnered him a million rep points, including the complexities of timezones (very Y2K problem). Plus, Matt DeBergalis from Apollo GraphQL jumped on to to talk the evolution of API orchestration and what it means for modern development stacks. And if you too are pondering about those peskily unreliable "dream machines," you'll be excited to know there are ways to base those dreams in reality. XX And in a world of unsimple things, we'll leave you with a few overly simplified answers: Is using your brain REALLY better for coding than an LLM? (Yes.) What kind of exception should you be throwing in C++? (Trick question, you shouldn't all the time.) Will T.S. Eliot get me fired from Apple? (Almost.) If you want to know more, the full answers are down below. XX
This week, we're focused on problems. Secretly, this is also a solution week, because if we just piled on with new problems, geez, what a bummer email that would be. We all know data is becoming a bigger engineering problem, so the podcast talked to folks at Tobiko Data about how they are improving the data engineering toolbox. For those of you in hardware, you know the troubles that tariffs have caused—our other podcast this week talks to MiniProto, a company doing hardware prototyping in the US. XX AI (thought you'd escape it this week, eh?) claims to be a solution, but it often comes with its own problems. We chatted with Vish Abrams, chief architect at Heroku, about the problems with software that vibe coders aren't thinking about. For those of you adopting AI at work, we review our Leaders of Code conversation with the CTO of Cloudflare to figure out how to overcome AI adoption and implementation problems. XX Okay, enough with the problems already. Time for answers! Why were players singing in a NASA MMO? Whose fault are radio buttons? How can I apply to more jobs, theoretically getting me one job? How can we distinguish these smart robots from people on the internet? FYI, these are questions—answers are in the links.
This week, we're thinking about the changes taking place within venerable institutions. We talked with the Chair of the OpenJDK Governing Board, Georges Saab, about what the past five years have meant to a 30-year-old language. The biggie: new releases every six months with whatever's done instead of waiting to shove a massive feature-fest out every few years. Over on the Leaders of Code podcast, our CTO talks with Cloudflare's CTO about how the internet—and their jobs making it a better place—are about to get a little cray-cray in the next few years. XX We're also thinking about how to keep making things better, now and in the future. We've got a round table podcast with Stack Overflow's Trust and Safety team about how they put out dumpster fires that are inevitable in any user community. For those of you in the Zoomer generation (or trying to deal with them spotting bad vibes), we've a fantastic look at how this generation learns differently and how AI fits there. XX Fret not, pretty babies, we have a fat stack of fun facts for you to trot out at the next "Kid from Jerry McGuire look-alike contest." Did you know giant snails could have sieged medieval castles? But did you know Marvin Minsky created a framework for thinking about AI agents in 1986? But, but did you know you can spot a bad workplace pretty fast? Bonus: Yes, Mr. Goodwillie, we have your 1987 letter!
Before we get into the goodies, please take a look at our Developer Survey for this year (link in the footer). It's an annual tradition that builds on the results of previous years to track the tech, habits, and opinions of software technologists. There's more in there about AI this year, but frankly, there's been more AI this year too, and we want to know if it's becoming a part of your workflow or if it's just another hype cycle. XX Speaking of AI, we've got two conversations about AI and an article about the data behind it. First, we chatted with the DeepMind team about Gemini 2.5, that plucky little GenAI tool that's barnstorming all the leaderboards. On the other hand, we've got a chat with mad scientists chaining together open-source models, compute be damned. If you're in the model training biz, we'd like to talk to you about diversifying your datasets. XX Now for the newsletter equivalent of speed dating. Could Leia have saved all those Bothans from doom? Is your List of Despair really about the list? Could you maybe go a little easier on the reauth? And rewiring your brain with jams, bops, and power ballads.
This week, we're thinking safety first. Hear about how to defend your mobile app when the attacker controls your compiled code and the device it runs on. Find out how time travel debugging works in a deterministic simulation. Check out the banking engineering org that went all in on serverless. And a safe bet: the Developers Survey is open for your input. XX You'll still find a little bit about AI here too (certainly more in the links). Read about how real businesses are integrating AI agents into their workflow and getting those sweet sweet productivity gains. If you're out there building AI on Snowflake, we've partnered with them to make our public site data easier to integrate for everyone. XX As promised, there are a couple of dueling AI takes in the roundup. "AI skeptics are crazy" vs. "no measurable productivity gains from AI". Two takes enter, a million takes leave! Other ponderables include the perils of a Magic: The Gathering monoculture and a how one human learned the frustrations of machine learning.
And we want you to show us! The 2025 Develop Survey is now open, and we couldn't be more excited (for health reasons)! What are you and your fellow software engineers talking about? Are you talking about how the future of RAG systems is agents? Or how your favorite CRM system is building agents? We sure are! XX Okay, maybe you're not talking about agents. We've got other conversations and articles for you. Want to find out why understanding AI limitations will help you succeed? Or maybe you've got an e-commerce front end and you want the dish on the latest improvements. Perhaps you're one of those door-number-three types and would like some TypeScript-powered software to explain your gut biome. XX Maybe, just maybe, you're here to scroll through the questions and links that make you thinks. Check out how folks are handling AI-based cheating in workplaces and schools. Get boggled by the maximum number of cube faces you could see at any one time. Marvel at the ways that hawks are using human tech to hunt.
We're rocking the mad scientist vibe this week, experimenting in our labs (which sadly do not have enough flasks and beakers). This quarter's CEO update lays out how we're tinkering and toiling to evolve into every developer's third screen. From our researchers, we've got a roadmap update, which dives into the specific experiments that we're running and what we're trying to learn. You can even check out one of those experiments now: Coding challenges are live! XX Turns out that we aren't the only ones trying new ways of doing things. On the podcast, Ben and Eira talk to Christophe Coenraets, SVP of Developer Relations at Salesforce, about how AI agents are changing the CRM game. When you're talking about experiments, you need to talk about safety, too. Our second podcast this week features a conversation about durable execution and making sure your microservices aren't playing in hardcore mode. XX Enough about us—what's everybody else talking about? Well, some folks are looking at the ways that AI is changing how software is being built, while others are saying that real productivity gains come from boring old processes and tech. Yet other people don't want to leave their oven on for the 15 minutes it takes them to prep the recipe. Productivity comes in all sorts of flavors.
Sorry for the jump scare in the title, folks, but this week, we're taking big picture looks at AI and the software industry it supports (or destroys). On the blog, we have a look at the big bets being placed on AI companies and how this could shake out in the long term. For those big companies gobbling up small AI firms, are they ready for AI in revenue-generating production? We chat with Maryam Ashoori, the product leader behind IBM's watsonx. If you're looking to hear about something more fundamental, check out our podcast conversation with self-confessed observability nerd Henrik Rexed on breakpoint debugging in cloud-native applications. In the age of moving fast, does velocity actually measure an org's impact? We expand on our latest Leaders of Code podcast to explore real measures of development work. More? Well, our organic artisanal web crawlers have found a slew of lovely links. Is it true that a cat has the IQ of a toddler? How can you scale an impenetrable wall of jargon? What happened when Iceland moved to a four-day work week? Does your system need all that complexity or is it just complicated?
Here at Stack Overflow, we try to keep our community central to what we do. We've got a bunch of products, sure, but the public site has been the core, and we're kicking off efforts to align our brand—the idea of us we present to the world—with all the things we do. Check out the blog and meta post announcing it. To stick with the community theme, our community management department is giving you a peek into what they do all day. XX Besides that, we've had a busy week. Check out why Algorand thinks TypeScript is the best way to build on blockchains. COO of LinearB Dan Lines talks with Ben Matthews and Ben Popper about moving from development velocity to impact. The SO podcast learns about orchestrating microservices. For all you robots afraid we'd skip AI this week, worry not: we chat with Abby Kearns of Alembic about using AI and contact tracing techniques to find which 50% of your marketing budget is worthless. XX Around the network and the web, information is popping (do the kids still say that?). Is AI pushing people into messianic delusions? Do the Middle Earthlings know that Frodo actually botched the job? What do you call those without sin? And did network science predict the next pope?
Everybody talks about GenAI in terms of prompts and responses, but we're going to focus on the infrastructure and protections that make it happen. In the last of the HumanX podcasts, Ryan chats with SambaNova CEO Rodrigo Liang about reimaging the chipsets behind AI. We have a Q&A with Jack Berkowitz of Securiti about the liabilities—legal and otherwise—posed by bad data. We also chat about GenAI and the other kind of infrastructure—the physical kind used by utilities—with Greg Fallon of Geminus. xx We're also honing in on innovation itself, that mysterious X-factor that turns problems into solutions. Cameron Wolfe shares his take on GenAI's OG innovation, self-supervised learning, and why it made such an impact on natural language processing. In our latest Leaders of Code blog post, we expand on our conversation with Hilary Lanham, Head of Engineering Excellence at Lloyds Banking Group, to look at the ways that innovation puts money in the banks. xx Did you hear that the electrical grids of Spain and Portugal were offline for a whole day? Take a dive into induced atmospheric vibration, the suspected cause, in our Questions section. Plus! The grind of performance optimization! Bot farms shaping the narrative! And war! What is it good for besides technological advancements?
Complexity is one of those things that computer science and software engineering has sought to reduce over time. No, don't laugh, I'm serious. While new techs may come in and make our lives messier, soon the industry turns its mind towards simplifications: standards, reusable components, platform-independent solutions, and so on. --- We've got a lot of conversations for you along those lines. In the latest Leaders of Code podcast, Ben Popper chats with Prashanth Chandrasekar, CEO of Stack Overflow, and guests from Lloyds Banking Group about how to simplify and standardize a complex engineering org. On the main podcast, we have chats with the CEO of Dataiku about simplifying the data to genAI pipeline, and with Laly Bar-Ilan, Chief Scientist at Bit, about how they are using genAI to create reusable components, not code. --- And in the questions and links, we're dropping knowledge you didn't even know you needed. Is women and children first a discriminatory policy? Can you host your blog on a Wii? Are millennials also flat earthers? Would you buy a computer from a raccoon?
Every other week, it seems like somebody has a new AI-related take or buzzword, and the rest of us have to sort out whether it's worth learning about. Like agents; what the heck are they? Fortunately, we dug into that very question to see what lies beneath the hype. Agents, like a lot of software, help automate tasks—we're got a look at how much you can automate sysadmin tasks. Then, with all these AI agents running rampant in your system, you need to think about scale—in our latest Leaders of Code, Recursion's VP of engineering Maureen Makes does just that. xx As we've often promised and occasionally delivered, we have more for you besides AI. In the latest We'll Be In Touch, Kyle sits down with former Stacker Michael Stum to chat about WebAssembly, how he learned to code by tackling Stack Overflow questions, and more. Speaking of Stack Overflow (aren't we always?), our private version, Stack Overflow for Teams, has a new release for Enterprise users. xx Okay readers, lightning round <airhorn>. How do you tell HR that you're quitting because your desk is terrible? Why does Discord see through time? What can you learn from the failure of a startup? Will the CVE database disappear when its grants run out?
No matter where in the world you live, this newsletter is hitting your inbox free of tariffs. I mean, it's always free, and we've got some excellent freebies for you. We dropped the results of our survey on open-source AI, and it's got plenty of insights into how developers of all experience levels view the rise of open-source LLMs like LLaMa and DeepSeek. -- On the podcast side, we have two conversations recorded at the HumanX conference and one new Leaders of Code episode. We chatted with Amr Awadallah, CEO of Vectara, about how AI's going to give us that "I know kung fu" moment sooner rather than later. We serve up a conversation with a repeat customer, Tomasz Tunguz, about his theories (and his Theory Ventures) about the AI landscape. In Leaders of Code, our own Ellen Brandenberger chats with Maureen Makes, VP of Engineering at Recursion, about whether you can move too fast with AI. -- Did you know that the #! is called a shebang? Me either. Find out how it works with the Linux kernel in our links section. Want to know how dang wrong Jules Verne (and, frankly, everybody else in 1865) was about the Rocky Mountains? Check our questions. Dear reader, there's even more, so I'll get out of the way of your scroll wheel.
Hey, have you all heard of AI? I kid, I kid, but it's popular for a reason. This week, we cover all sorts of angles on the topic du jour in today's software industry. We chat with the CEOs of HumanX and Crunchbase about the state of the AI industry. We chat with the field CTO of GitLab about how AI best supports learning. We chat with folks at the HumanX conference about audio models and LLM governance. There's more to AI than just LLM, and we've got leads on that too. There's info on the new role web data will play in LLMs. Our friends at Intuit talk about an old-school AI technique they use to manage the NP-hard problem of scheduling appointments at scale. In other links, we've got the usual roundup of delicious info. Ponder the math of roasted meats. Remember one of the code breakers who helped stop the Nazis. Learn about the movement to develop web apps for local-first execution. And argue whether there were CRT displays that could refresh faster than 60Hz.
We talk a lot about AI and its effects on coders. The reality is that the keywords and syntax many of us use to activate software magic might not be as important in an AI future. In the latest Leaders of Code podcast, our CEO Prashanth Chandrasekar talks to GitLab Field CTO Lee Faus about why critical thinking might be the new must-have developer skill. There are plenty of harder tech skills that AI requires. Vector databases and embeddings have been the foundation of many GenAI advances, and we talk about it with Pinecone CEO Edo Liberty. On the other end of the AI spectrum, we talk about the thing that everybody is talking about now: agents. Deepak Singh, VP of Developer Agents and Experiences at AWS, breaks down whether these plucky little robots are ready for prime time. Meanwhile, have you ever thought about rewriting the C++ standard library? Me either, but we have a link to someone who did. Plus Go slices, random PC crashes, homophones gone wild, and translating an artistic perspective into math.
We just flew back from the HumanX conference, and boy are our arms tired...from holding a microphone to interview so many interesting folks about AI. We've dropped the first interview from the conference, and forgive us, it's a self-link. Our CEO Prashanth Chandrasekar talks all about our place in the AI landscape and what we found interesting over the three days at the conference. -- Naturally, we're sharing a greater bounty in this newsletter. Want to know about building a mobile framework, tracing errors with AI, and the best way to fine-tune your data strategy? We've got you covered. -- So what's the rest of the Stack Exchange network, perhaps the world at large, talking about these days? Trained crows, the speed at which the days get longer, and the definitions of font and printer. As a closer to this meal, please enjoy a little dismantling of the worry that AI and vibe coders are going put you terminal jockeys out of business.
We've been in full swing lately, and we have a ton of exciting things for you to check out. First up, a look at how developers caught between caring for kids and aging parents are managing to keep focused on their work and use software to help with that. Next, take a look under the hood of our sweet new Question Assistant. Finally, our very own Eira May got up close with agents at TDX 2025, and she's unpacking what all that means for developers. / For those who prefer to listen, we've got a ton of goodies. Check out our very first episode of Leaders of Code, a new podcast that pairs one of our leaders with a tech leader at another organization. In the good ol Stack Overflow Podcast, we've got conversations about startups trying to combat climate change and a terminal built to use AI to bypass man pages. Curious crawlers will find a collection of clickables. Want to destroy the Milky Way? We've got you. Hide data using getters? Done. Build a calculator entirely out of wood? You absolute lunatic, we can help.
As you read this, your humble narrator will be at the HumanX conference talking to leaders in the world of AI. Whether we like it or not, AI is becoming a key part of the technological landscape. This week, we feature two conversations with people building real AI solutions: optimizing models for edge devices and using specialty models for translation and beyond. We do have other non-AI material for you. The podcast welcomed a blog contributor for a conversation on secure coding, what it takes to secure a Canadian election, and the legal ramifications of being a lousy CISO. Plus, our Teams product has shiny new features and we're excited to tell you about them. Never fear my intrepid knowledge explorers, there are plenty of goodies from the Stack Exchange network and beyond! Is the Earth dodging the Moon? Can you learn HTML and CSS from a game? Can a catch-all email domain be abused by bad actors? Can SQLite be abused to handle concurrency? And more.
Here in the Northern Hemisphere, we are on the cusp of Spring, so we're thinking about the growth and change coming in the future. The weather is getting slightly warmer, brave flowers are budding, and the snow is melting. For Stack Overflow, we're thinking about how we can grow and change to keep our public community vibrant in the face of an AI present. To that end, we had a Q&A with some of our leadership team over YouTube. If you missed it, we've got a quick recap that touches on all the possibilities that we're mulling over. Of course, we're also talking about AI: how testing can benefit and the variations of LoRA. And if you want to share your views on AI instead of reading and hearing ours, we're running a survey on using open-source AI. We also have a collection of curated curiosities for you, surfaced by the Stack Overflow team and Cassidy Williams. If you want to build a pipe organ, get your tiny gecko friend honest work (or poach them from their current position), or configure Git the way core Git developers do, we've got you covered.
This week: keeping software development both fast and safe, breathing something other than air, and how your morning coffee changes your brain.
This week: we're gearing up for an AMA with our CEO, growing paranoid over our Linux kernel's knowledge of Microsoft tech, and celebrating the return of the pebble.
This week: Feature flags in theory and in practice, the world’s best mirror, and biohacking a broken heart.
This week: The organizations to which Stack gave back in 2024, what you can do when your open-source plugin is repackaged as commercial code, and the UX of login codes.
This week: The oft-forgotten soft skills developers need, decoding macrodata refinement, and the challenges icon-only buttons represent for screen readers.