Normal view

There are new articles available, click to refresh the page.
Yesterday — 22 July 2026Tech

Stop Overengineering Your Agent Harness

22 July 2026 at 11:58
The following originally appeared on Hugo Bowne-Anderson’s Vanishing Gradients Substack and is being republished here with the author’s permission. The conversation around harness engineering is dominated by problems from coding and personal agents such as OpenClaw, but most agents are simpler. Builders should avoid over-engineering for capabilities that newer models may absorb anyway, the “Kirby […]

Managers Are Not Overhead: They Are Infrastructure

22 July 2026 at 06:42
Managers have been disproportionate casualties of the rolling waves of post-COVID-19 tech layoffs that started in late 2022. Popularized by large companies such as Meta, Google, and Amazon, phrases like “flattening the org” and “reducing bureaucracy” are now synonymous with thinning the management layers that ballooned during the 2021–2022 hiring sprees. Retrospectively, such flattening can […]
Before yesterdayTech

Zero to Agent in 30 Minutes: Build a Workflow Agent with John Berryman

20 July 2026 at 16:50
We kicked off Zero to Agent in 30 Minutes this week with guest John Berryman, an AI consultant and contractor for Arcturus Labs. John has spent the past several years building AI products and consulting on how teams put them into production. He set the stage by defining an agent as a large language model […]

The Tokens You Can’t Wait For

20 July 2026 at 06:59
Somewhere in a Singapore data center, a bank is paying for eight H100s that spend most of the night waiting. The cluster was bought for good reasons (discomfort with customer documents leaving the building, a strategy team’s aversion to lock-in), so the bank secured its own sovereign compute. Now the finance team is asking why […]

This Week in AI: A First for Agentic Ransomware

17 July 2026 at 11:54
Christina Stathopoulos, the data and AI evangelist behind Dare to Data, continued her run sorting the week’s most impactful stories into a handful of themes we’ve been watching play out over the past month: more firms investing in the compute AI runs on, more concerns about who controls a model’s borders, and more AI-generated code […]

Coding Was Never a Bottleneck

16 July 2026 at 07:12
AI has taken software development by storm. Between the two of us, we build products for software engineers and consumer products for millions of everyday users, so we have skin in the game. We want the AI productivity story to be true. More output, tighter timelines, happier and more productive engineers. Who wouldn’t? But when we […]

Don’t Neglect the Operational Groundwork

15 July 2026 at 13:00
Autonomous agents are moving faster than the field’s ability to govern them, and catching up requires more than better prompts or bigger sandboxes. At O’Reilly’s recent AI Superstream focused on OpenClaw and the broader ecosystem of locally run and self-hosted AI agents, five speakers, each working at a different layer of the stack, explored patterns […]

The New Software Lifecycle

15 July 2026 at 06:54
The following article originally appeared on Addy Osmani’s blog and is being republished here with the author’s permission. I cowrote a Google whitepaper about how AI is changing the software lifecycle. I’m not going to summarize the whole thing. Instead, here are the handful of ideas in it I think actually matter, plus six figures […]

The Open Source Agent Toolkit in 2026

14 July 2026 at 06:57
The following article originally appeared on Paolo Perrone’s Substack, The AI Engineer, and is being republished here with the author’s permission. You spent three weeks shipping an agent. It worked in the demo. Then production hit, and you realized the framework you picked has no checkpointing, the memory layer is a flat vector dump with […]

This Week in AI: Chips, Checks, and Changing Jobs

10 July 2026 at 12:04
This week data and AI evangelist Christina Stathopoulos returned for a solo news briefing. Instead of exploring one or two topics in depth, Christina sorted the week’s headlines into a handful of threads: advances in physical hardware to keep up with AI demand, the widening reach of government oversight into frontier model companies, and a […]

Opinion: The WALL-E Economy

10 July 2026 at 09:40
A scene from Pixar’s 2008 film WALL-E. (Image: Disney/Pixar)

[Note: Armon Dadgar, a UW computer science alum, was co-founder and CTO of HashiCorp, a cloud infrastructure automation company that was founded in 2013, went public on the NASDAQ in 2021, and sold to IBM in 2025. In his role, he spoke with thousands of companies adopting cloud across a wide range of industries, giving him a unique perspective on technology adoption.]

Pixar released WALL-E in June 2008, almost two weeks before Apple launched the App Store. The film follows WALL-E, a robot left behind to clean up Earth after it becomes an ecological wasteland no longer suitable for life, stranding humanity out in space. The remaining human population is entirely moribund in hoverchairs and completely immersed in a digital reality.

For many years, I’ve jokingly called it a historical documentary from the future and with each passing year we seem to get closer to its dark prophecy. Today, we live in a “WALL-E economy” with apps and services that cater to convenience and human vice, but with an increasing toll to our mental, physical, and emotional wellbeing, as well as our environment. The growing capabilities and prevalence of AI risks accelerating those trends, and moving us further towards the WALL-E dystopia.

The faustian bargain of WALL-E is that we willingly trade comfort for everything, including our free will. We live a bovine existence, where we are endlessly fed content, told what to wear, what to buy, what to think, and how to vote, and in exchange we are kept safe and warm in the proverbial womb. We get our dopamine, but we never ask at what cost.

Far from being hypothetical, we can see many of these same tradeoffs in the most popular services today. The average American spends around 2.5 hours per day on social media, and for Gen Z it’s even worse, at a stunning 5 hours daily. It feels costless to scroll through memes and share posts with friends, but these services have stolen time from physical activity, connecting with friends, engaging in hobbies, and other activities that foster connection and meaning. We have a crisis of teenage mental health, a loneliness epidemic, rising political extremism, and a “friendship recession.” While correlation is not causation, it’s reasonably safe to admit it’s the phones.

Outside of social media, we have many services of convenience, such as DoorDash, Uber Eats, and Instacart, which are primarily used for food delivery. While you might think they are “luxury services,” data shows that usage is broad across income levels, and in fact disproportionately used by those who can least afford it, even trapping users with “Buy Now, Pay Later” services. The immediate convenience is undeniable, but so is the financial burden to users, damage to restaurant culture as they are forced to adapt to delivery services, and the growing unhappiness that comes with solo dining.

The WALL-E economy often pairs delivering convenience with packaging vice under a shiny application layer. Sports betting services like DraftKings, prediction markets like Kalshi, and retail options trading on Robinhood are prime examples. They are all gambling packaged up and made presentable. Chesterton’s fence reminds us that if we’ve discouraged gambling for millennia, there might be a good reason for it. Unsurprisingly, making these services available 24/7 and putting them in our pockets has led to a rise in addiction, bankruptcy, and suicide.

If the smartphone enabled the WALL-E economy, AI is going to supercharge it. AI is accelerating the ability to collect and analyze data, to highly personalize, and to target algorithmically with precision, with the intent to influence user behavior. This threatens to make the WALL-E economy both smarter and more harmful.

Targeting people who are already lonely and alienated, virtual dating services are one of the most pernicious use cases for AI. Users are willing to believe they’ve found meaningful connection with an AI while getting used to the “frictionless” interaction of chatbots makes the nuanced and ambiguous world of real people that much harder to navigate. It’s clear the cure is worse than the disease, as this only increases social isolation, leading to depression and suicide.

AI will also enable more sophisticated applications of surveillance capitalism and dark patterns that aim to manipulate user behavior. AI-driven customer profiles will determine if you are a price sensitive shopper and use that to inflate prices or apply surge pricing. Betting apps can detect when a frequent gambler hasn’t placed bets recently and give them free credits to lure them back along with a 24/7 AI-bookie that can discuss and encourage bets.

The AI-slop future of social media is clear as Zuckerberg is walking back from the metaverse and pivoting to AI. Social media platforms historically needed users to actually create content, which they could promote based on user interests. The need for content producers vanishes if content can be AI generated, perfectly tailored and personalized, endlessly. This future might increase user engagement, but would likely exacerbate the problems of isolation and alienation that we already face.

We can wax nostalgic for a simpler time, but there is no path back to before smartphones, social media, or AI. We can’t wish it away, and there are staggering amounts of capital being deployed to extend AI to every corner of the economy. While Silicon Valley might argue for “technology inevitability” and the notion that technological progress is inevitable and good, this ignores the fact that people still have agency. Most technology is inseparable from a set of social and political questions, and unfettered use is not inevitable. The recent order banning Anthropic’s Fable and Mythos model is a prime example.

In the movie, WALL-E is given the sisyphean task of cleaning up a ruined Earth. Despite his programming, he imagines a better world and rejects the inevitability of his destiny to fight for something better. Today as we consider the AI-supercharged version of the WALL-E economy, it’s not surprising that most people aren’t enthused. We are suffering from a failure of imagination for what a better future could look like instead.

The rise of “friction-maxxing” as a new trend that rejects convenience as an end to itself is a start. It’s part of the growing recognition that humans need to derive meaning from our lives and work. The growing discourse around the perils of convenience, along with individual behavior changes, help to shift the Overton window. Changing hearts and minds is a crucial step towards new regulations, which are needed to solve the social challenges more broadly. This is a slow process, but we can look to cigarettes as a good historical analogy.

Cigarette companies had a clear incentive to market and sell aggressively, and made a best effort to hide the health impacts of their products, similar to the companies powering the WALL-E economy. Eventually, it became clear that cigarettes pose a major health risk, both to the direct users and the second-hand bystanders, which prompted social changes in how they were viewed and ultimately political changes in how they are regulated. As a result, there has been a dramatic reduction in their usage today.

For the modern WALL-E economy there is no perfect, singular regulation. It requires a democratic process to balance mitigating the harms as we better understand them, with individual rights and autonomy. For digital platforms, countries like Australia and the UK are leading the way in banning the use of social media for children and teenagers, in recognition of the harm they cause. Utah is taking action against “prediction marketplaces” in recognition of them being effectively online gambling. This type of regulatory change is important to solve societal problems but takes time, especially with highly concentrated interests lobbying against a disorganized public.

Beyond just waiting for improved regulation, we can reclaim our individual agency by living with more intentionality. In my personal life, I’ve looked for opportunities to host more regularly, whether a small dinner or a large party to create and foster connections. I’ve worked to reduce the “information noise” by disabling most notifications, limiting time on social media, and spending more time reading on my Kindle distraction free. I consciously avoid eliminating all daily friction to provide a healthy level of resistance. None of these are huge changes, nor will they solve the broader problems, but they help to reshape our environment to avoid depending on willpower alone.

Changing our personal behaviors can help to insulate us from the WALL-E economy, but that isn’t enough. Taxing the profits of AI companies to provide UBI to fund “bread and circuses” couldn’t be more WALL-Esque. We should look to the techno-optimism of the Jetsons, where technology empowered humanity, and build towards a future we are excited about, rather than one we merely accept as inevitable. WALL-E is incredibly dark for a children’s movie, but the struggle of the character ultimately gives hope to a more optimistic future, and brings to mind the old boy scout motto: try to leave this world a little better than you found it.

Special thanks to Josh Kalla, Behzod Sirjani, and Kevin Fishner for their feedback on this post.

Prompt Injection to Data Exfil in 3 Hops

10 July 2026 at 06:45
The incident that should worry you makes no destructive call. Nothing is deleted, nothing crashes, no alert fires. An employee asks an agent to summarise a customer ticket; the agent does exactly that, the user gets a useful answer, and somewhere, in the same second, a customer record leaves the cluster over an ordinary HTTPS […]

Why the video game industry may be sliding toward its next big crash

9 July 2026 at 17:47
4 generations of Xbox hardware. (GeekWire Photo / Thomas Wilde)

Commentary: The last couple of weeks have served as a capstone to what’s become a bad few years for the international video game industry. Now it appears the larger sector is headed directly into a significant crash, as several unsustainable practices all seem to be approaching a crisis point at once.

The first and most obvious issue is the ongoing component shortage. Due to the rush to build AI data centers, both RAM and solid-state drives have risen dramatically in price in 2026, with analysts forecasting that costs might not settle back down until at least 2028.

Both the PlayStation 5 and Xbox Series X|S are at the point in their life cycle when they’d ordinarily be declining in per-unit costs as the technology matured. Instead, both Sony and Microsoft have raised console prices multiple times this year due to the high demand for parts.

This would ordinarily be a great time to get into video games, as we’re almost six years into the current console generation. Instead, it’s one of the worst. The base PS5 and Series X are about as expensive as they were at launch in November 2020, and building a new gaming PC right now can be costly.

The component crunch also harmed the debut of Valve’s new Steam Machine, which officially launched late last month with a starting MSRP of $1,049. Valve, based in Bellevue, Wash., was forced to offer the new hardware at a significantly higher price than planned due to the difficulty in getting components.

That’s been reflected in its early reviews, with many outlets noting that the Steam Machine’s current price doesn’t match its power. At $700, the Machine would be a great gateway product for PC gaming, the way the Steam Deck was, and a genuine competitor in the console field, but a $1,049 price tag makes it an expensive curiosity for financially secure gadget-heads.

Another bad sign came from Sony’s recent announcement that it would sunset physical media for the PlayStation platform by 2028. This decision, which allegedly took many of Sony’s publishing partners by surprise, has serious knock-on effects for collectors, historians, developers, and most prominently consumers.

Sony has already caught one lawsuit over alleged market exploitation on the PlayStation Store, and that was a few days before it announced it wants to kill discs. An all-digital PlayStation library means that Sony would get to exercise full monopolistic control over pricing and access for every game it sells; licensing agreements mean that anything purchased on a digital storefront like the PlayStation Store is subject to deletion at any time without notice; and players wouldn’t be able to resort to any of the usual cost-cutting measures such as bargain bins, buying used copies, or even trading games with a friend.

That suggests that Sony has decided its best path forward is to continue to extract money from its established audience, rather than to have more options in place for gaming on a budget. There are free-to-play games on the PS5, of course, but most if not all are cross-platform and/or designed as money sinks. Ask any parent whose kids accidentally ran up a big tab in Fortnite.

Sony’s PlayStation 5. (Sony press image)

If Sony has decided to end physical media, then it’s likely Microsoft will follow suit. While Xbox hasn’t mentioned its next-generation console, codenamed Project Helix, for a hot minute, it has been eager to get rid of discs since at least 2013. Some sources, such as Windows Central, allege that Xbox is already planning to do so.

(Meanwhile, Nintendo is likely to do its own thing. While Nintendo has been forced to raise the price of the Switch 2 alongside its competitors, it has offered no sign that it plans to stop selling game cards or Switch cartridges. In an uncertain world, Nintendo can be relied upon to only ever follow its own peculiar instincts.)

This sets up an early look at the environment that surrounds the 10th generation of console hardware. If both Sony and Microsoft stick to traditional timelines, we’re likely to start hearing more about the PlayStation 6 and Project Helix over the course of 2027, with launch in holiday 2027 or 2028.

If they do launch along that timeline, then it’s difficult to see how either system will retail for less than $1,000, since the storage and RAM supplies will still be constrained by that point. That automatically prices most of the potential audience out of the market. Once the starting costs hit the four-digit range, a console stops being a hobby or a toy for children and becomes an expensive extravagance. (As a general rule, you probably don’t want your console to cost significantly more than the TV you’re attaching it to.)

Further, it’s arguable that neither the PlayStation 5 nor the Xbox Series X|S have really hit their potential. Sony has famously squandered much of this generation on a largely abortive pivot to games-as-a-service, while Xbox has often seemed more interested in laying off developers than actually making or marketing games. The 9th generation of consoles has had a few big hits, but it’s mostly despite itself.

Not only is there likely to be limited demand for the 10th-generation PlayStation or Xbox, but neither of them actually seem necessary. The only reason to make them is for a brand refresh, and that’s got nothing to do with consumers.

Microsoft, following its acquisition of Activision Blizzard in 2023, is currently the second largest game developer in the world, while Sony dominates today’s console market. These two companies influence much of what happens in the modern video game industry, and as of right now, both are apparently determined to do the most short-sighted thing possible at any given time.

Sony has decided that only part of its audience actually matters, while Microsoft seems to be saddling Xbox with unrealistic expectations, possibly to justify its eventual sale or shutdown, and is ignoring at least one organized boycott.

Reggie Fils-Aimé (center) leads a roundtable discussion of Xbox architects to celebrate the platform’s 20th anniversary in 2021. Left to right: Robbie Bach, Ed Fries, Fils-Aimé, Peter Moore, Bonnie Ross. (Microsoft Alumni Network)

Whenever the video game industry undergoes any kind of significant disruption, someone somewhere always asks if it’s the start of another “Crash of ‘83.” This is usually hyperbole, but it’s hard not to see the parallels between then and now: the video game market is flooded, there are few true exclusives left outside of Nintendo, many members of the gaming audience buy as few as 2 games a year, and the end of physical media will end both retail support and much of the casual audience.

This is unfolding as a slow, years-long plummet rather than the comparatively sudden shock of ‘83, but a crash is a crash. It’s avoidable, but it would require a massive, simultaneous course correction from several of the largest entertainment companies in the world.

That being said, it’s unlikely that video games as a medium are facing any kind of existential threat. Nintendo, as noted above, is well-positioned to ride out any potential problems with the larger market, PC gaming is hanging on, and the mobile sector is actually having a sort of quiet renaissance right now. There will still be video games to play in 2030, barring some larger disaster.

If there’s one big opportunity here, it’s that many of the major players in the games industry have either voluntarily abandoned the market for budget gaming or have been forced out by component costs. Some of the biggest hits of the 2020s to date, such as Vampire Survivors, Among Us, Lethal Company, and Balatro, are cheap, retro-styled games designed to run on almost any hardware, from a PlayStation 5 to your 4-year-old tablet.

The best step forward for mainstream gaming, then, might actually be to take a step back, in a similar way to projects such as Panic’s Playdate retro handheld (still going strong 5 years later) or Seattle’s Tin Can, seeing success with its land-line phones for kids and families. Chasing bigger games, higher frame-rates, and more realistic graphics for 30 years has gotten us here, up to the edge of a second major crash, while thousands of people log on every day to play games that could be run on a particularly big potato.

Instead of rushing into the 10th generation, the solution now might be to think simpler and cheaper, making smaller, more focused projects rather than the 5-year moonshot of a typical AAA game. Otherwise, mainstream video games may end up like Western comics: increasingly expensive options presented to a shrinking handful of fervent fans.

AI Enthusiasts Are in a Race Against Time, AI Skeptics Are in a Race Against Entropy

9 July 2026 at 07:00
The following article originally appeared on Charity Majors’s Substack and is being republished here with the author’s permission. I recently attended a talk where one of the presenters made some pretty…astonishing claims about what they had achieved by the pure, uncut power of vibe coding. Difficult engineering problems solved, backlogs cleared. Rewrites that would have […]

Why AI Coding Agents Still Need Clear Specs

8 July 2026 at 07:03
The following article originally appeared on Markus Eisele’s newsletter, The Main Thread, and is being republished here with the author’s permission. There’s a mental model spreading through the developer community right now that goes something like this: Agents are smart enough to figure things out, so heavy upfront specification is bureaucratic overhead you don’t need […]

Ordinary Engineers, Not Heroic Inventors

7 July 2026 at 07:38
In the 1980s, Japan led the world in semiconductors, consumer electronics, and computer hardware, the industries everyone assumed would decide the next phase of economic power. Japan won them and still did not overtake the United States in the information revolution that followed. Jeff Ding, a political scientist at George Washington University, opens his book […]
❌
❌