Isaac, 33, has been a mid-level software development engineer at a Big Tech firm for four years, and noticed entry-level job postings dropping at his workplace at the start of 2025. The work, however, didnt vanish with them. Tasks once handled by junior engineerslike writing and testing code, fixing bugs, and contributing to development projectswere absorbed by senior staff, often with the assumption that AI would make up the difference.And while AI has sped up the velocity of shipping code and features, there are fewer people to do tasks like designing, testing, and working with stakeholders, which AI has zero grasp on. The cracks have been hard to ignore. Seniors are burning out, and when they leave, theres no rush to replace them, because the AI will do it! Isaac says. Worried that hell become the next strung-out senior, hes looking for his exit, ideally at a smaller tech firm. (Isaac spoke to Fast Company under a pseudonym to avoid possible retaliation.)
The shift is striking, given how recently corporate America was courting Gen Z with fanatic fervor. Organizations raced to prove they understood younger employees. They flooded LinkedIn with thought leadership on the multigenerational workplace of the future, and retooled benefits programs to include wellness stipends and mental health days. Reverse mentorship programs, through which younger employees share knowledge and perspectives with more senior colleaguestouted by companies like Target, Accenture, and PwCpromised to give junior employees a voice in shaping culture and strategy. Some firms even brought Gen Z voices into the boardroom.Yet now, in the case of firms like Isaacs, entry-level workers, once heralded as essential to innovation and growth, are struggling to get a toelet alone a footin the door. Internships, starter jobs, and junior roles, the indispensable on-ramps to white-collar careers, have been evaporating for several years due to cost pressures and post-pandemic belt-tightening. Since 2023, entry-level job postings in the U.S. have sunk 35%, according to labor research firm Revelio Labs.The advent of AI is accelerating the entry-level apocalypse. Two-fifths of global leaders revealed that entry-level roles have already been reduced or cut due to efficiencies made by AI conducting research, admin, and briefing tasks, and 43% expect this to happen in the next year.
While theres steady hiring or even growth in the skilled trades, were seeing entry-level vacancies fall significantly in tech and customer service and sales roles, says Mona Mourshed, founder of the workplace development nonprofit Generation. Being in the business of training and placing people into entry-level roles, we find it deeply concerning. Graduates are clearly not okaybut neither are the companies that decided they could do without them.
AI at work: the supercar with no driver
The logic was seductive in its simplicity. Cut costs, move faster, shrink training budgets, let AI and a leaner workforce handle the rest. In reality, its producing something else entirely: flattened teams with little agency, endless cycles of rework, and exhausted senior employees juggling all task levels at once.
One redditor who posted about how their company has stopped hiring entry-level engineers, received hundreds of other responses as others chiming in with similar stories. One commenter noted: Not sure what the plan will be after the knowledge transfer is over.Isaac has watched this dynamic unfold firsthand. Leaders at his company see AI as a force multiplier, and are fixated on shipping features quickly. Isaac can see their point: [AI] can straight up write better, faster, more legible code than most developers, he admits. However, he points out, any seasoned engineer knows the hard part isn’t writing the code, its the design and testing. Yet, theres far fewer people to delegate this work to, so senior developers are left to do this on their own.
Compounding the problem is the fact that AI doesnt understand the problem its meant to solve. Left unchecked, it can go rogue. Isaac recalls multiple instances of chatbots deleting production stacksunpromptedbecause they couldn’t figure out how to solve an issue. Without an expert who knows how to prompt and guide it, AI is just a supercar with no driver, he says. The team has seen their workload steadily increase in line with automation, so the time savings it creates have had little impact. Many seniors have checked out, with several burned out engineers signed off for medical leave.
Research from the project management platform Asana underscores this growing “efficiency illusion.” While 77% of workers are already using AI agents and expect to hand more off to them in the next year, nearly two-thirds say the tools are unreliable, and more than half say agents confidently produce incorrect or misleading information. The result is time down the drain: a U.S. study found that employees are spending an extra 4.5 hours a week fixing AI workslop. AI can make work look faster on the surface, but it can also create a lot of cleanup workdouble-checking outputs, correcting errors, and redoing steps that were based on faulty information, Mark Hoffman, Asanas Work Innovation Lead, tells Fast Company. When something goes wrong, accountability is murky, he adds, and the responsibility often falls back on the employee to catch errors, explain outcomes, and manage the risk. Its driving up already record-high levels of burnout; 77% of knowledge workers say their workloads are unmanageable, and 84% are digitally exhausted.When errors slip through, the consequences are costly and embarrassing. Three-quarters of Americans report at least one negative consequence from poor AI outputs, including work rejected by stakeholders (28%), security incidents (27%), and customer complaints (25%). In October, Deloitte was forced to refund the Australian Department of Employment and Workplace Reltions after a report was found to contain AI hallucinations and workslop. In the past, newbie consultants would have handled tasks such as this. However, notably, Deloitte cut its graduate cohort by 18% and slashed hundreds of early-career roles earlier that summer.
The demographic time bomb
Not only are workloads increasing, by hollowing out their junior ranks, businesses are putting themselves squarely in the path of a slow-burning demographic time bomb as seniors begin to retire in record numbers.From 2024 to 2032, 18.4 million experienced workers age 55 to 64 with postsecondary education are expected to retire, but only 13.8 million younger workers (currently age 16 to 24) are entering with equivalent qualifications. Even in an AI-powered economy, where certain jobs will be automated, companies still need humans with judgment-, context-, institutional-, and sector-specific insight.
Yet plenty are making movesat least for todayto wipe out the training ground that turns beginners into experts.There wont be an endless supply of experienced hires to fall back on, so everyone will be fighting for the limited, increasingly expensive talent with domain expertise, says Cali Williams Yost, futurist and founder of flexible-work consulting firm Flex+Strategy Group. Companies have maybe five years to train younger workers to take over and gain the niche knowledge, so AI has something to augment.
Moe Hutt, an entry-level recruitment marketing expert and director of consulting at recruitment marketing agency HireClix, has watched clients scale back or abandon entry-level hiring, citing AI-aided workflows and economic uncertainty. Hutt points to the less visible fallout within organizations beyond damaging the talent pipeline. Its human nature to want to help, she says. When theres no release valve of training juniors, it creates friction everywhere.
For middle and senior management, delegating, teaching, and watching someone grow is a reward for the experience. Research consistently shows that sharing knowledge and mentoring improves motivation, boosts psychological well-being, and reduces burnout among experienced employees. With no one to train or teach, disengagement spreads, eroding a workforce where most people have already checked out.
Being AI-savvy and being prepared for the demographic cliff arent mutually exclusive. Organizations can build pro-worker environments where employees are augmented with AI, without hollowing out their future talent pipelines. PwCadmittedly, another firm which has been open about its cuts to entry-level recruiting, at least in the U.K.is experimenting with what that balance could look like by training junior accountants to become managers of AI. Entry-level employees gain early exposure to leadership and accountability, while the firm builds a cache of managers that are fluent in both human judgment and machine output. Its proof that efficiency and succession planning can coexist.
This matters because disappearing entry-level jobs arent just a problem for the corporate workforceit will be a societal crisis, too. A functioning society depends on younger generations steadily taking over from older ones. AI might be able to write the code, but without people trained to guide it, question it, and eventually replace their elders, there will be no one left to keep the lights on.
Until recently, Peter Attia was best known as a wellness influencer and a newly appointed contributor at CBS. He hosts a popular podcast, boasts more than 1.6 million Instagram followers, and wrote a best-selling book about longevity.
That image cracked this week when it was revealed that Attias name appears more than 1,700 times in the latest Epstein files release.
As the emails circulated on social media, longtime followers of his methods, along with medical professionals, reacted with outrage.
Peter Attia being Epstein’s concierge doctor is by far the weirdest crossover, one X user wrote. Another one X user quipped: Peter Attias stress level right now must be entirely undoing whatever longevity gains he has enjoyed.
Peter Attia being Epstein's concierge doctor is by far the weirdest crossover— AJAC (@AJA_Cortes) February 1, 2026
Others said the revelations were not surprising. The Peter Attia stuff is sad but not surprising, author Brad Stulberg posted on X. The entire health, performance, and longevity space is filled with narcissistic sociopathic grifters.
The Peter Attia stuff is sad but not surprising. The entire health, performance, and longevity space is filled with narcissistic sociopathic grifters.— Brad Stulberg (@BStulberg) February 1, 2026
Why are people surprised that someone who dropped out of residency, worked for an international consultancy firm, pretended to cry on TED stage, and charges over $100K/year for concierge “longevity” medicine…turned out to be unsavory? another medical professional wrote. Some of y’all will never learn.”
Why are people surprised that someone who dropped out of residency, worked for an international consultancy firm, pretended to cry on TED stage, and charges over $100K/year for concierge "longevity" medicine…turned out to be unsavory?Some of y'all will never learn.— Remnant | MD (@RemnantMd) February 2, 2026
Some zeroed in on new context around Attias own prior admission, detailed in his book, that he failed to be at his wifes side when his son fell seriously ill. As his baby son lay in ICU, Peter Attia told his wife that he was too busy with important work in NYC to fly home to see him, one X user noted. He was with Jeffery Epstein.
As his baby son lay in ICU, Peter Attia told his wife that he was too busy with important work in NYC to fly home to see him. He was with Jeffery Epstein. pic.twitter.com/lcfBMcaa5z— Dave @ Longevity Labs (@Dave_Longevity) February 2, 2026
The released correspondence shows Attia maintaining a friendly and at times flippant tone with Epstein. In one message dated June 24, 2015, Attia wrote: You the biggest problem with becoming friends with you? The life you lead is so outrageous, and yet I cant tell a soul In perhaps the most gratuitous email, Attia joked about the carb content of performing a sexual act.
In July 2016, Attia asked Epstein what he was doing in Palm Beach, where Epstein allegedly sexually abused underage girls during the 2000s. Guess, Epstein replied. Attia answered: Besides that.
Attias regular correspondence with Epstein continued years after Epstein pled guilty to soliciting prostitution from a minor. In a lengthy X post addressing the emails on Monday, Attia said he questioned Epstein about those charges and claimed Epstein grossly minimized them.
Attia has since stepped down as chief science officer of David Protein, the company confirmed Monday. CBS editor in chief Bari Weiss has also faced public calls to cut ties after Attia was named among 19 new CBS News contributors just days before the emails were made public.
Attia said on X that he never witnessed illegal behavior and never saw anyone who appeared underage in Epsteins presence. He added that he was never on his plane, never on his island, and never present at any sex parties.
The following email is what I sent my team last night. I sent a similar version to my patients, also. ***Youve put your trust, your credibility, and your hard work into what we have built together, and I take that responsibility seriously. You deserve a complete and honest— Peter Attia (@PeterAttiaMD) February 2, 2026
Other released emails show Attia saying that he goes into JE withdrawal when I dont see him and suggesting that he hoped to one day go to Epsteins private island. I need to visit some time, Attia wrote.
Want more housing market stories from Lance Lamberts ResiClub in your inbox? Subscribe to the ResiClub newsletter.
During the Pandemic Housing Boom, housing demand surged rapidly amid ultralow interest rates, stimulus, and the remote work boom. Federal Reserve researchers estimate new construction would have had to increase by roughly 300% to absorb the pandemic-era surge in demand.
Unlike housing demand, housing stock isnt as elastic and can’t quickly ramp up. As a result, the heightened demand drained the market of active inventory and caused home prices to overheat, with U.S. home prices in June 2022 sitting a staggering 43.2% above March 2020 levels.
Since that national boom ended in mid-2022, the housing market has been moving through a cyclical cooling phase and undergoing a period of recalibration and normalization after such a large burst.
Look no further than the share of U.S. homes that sold below their original list price, by year, according to a new Redfin report:
2018 > 62%
2019 > 64%
2020 > 55%
2021 > 38%
2022 > 42%
2023 > 54%
2024 > 58%
2025 > 62%
The share of homes selling below their original list price varies by region. Many Sun Belt pandemic-boom marketsparticularly across Florida and Texasare seeing the highest prevalence of homes selling below their initial ask.
By contrast, many Northeast and Midwest metros remain, relatively speaking, more resilient, with fewer than half of homes selling below list in several markets. Parts of San Francisco and San Jose have regained a bit of mojo amid the AI boom.
window.addEventListener("message",function(a){if(void 0!==a.data["datawrapper-height"]){var e=document.querySelectorAll("iframe");for(var t in a.data["datawrapper-height"])for(var r,i=0;r=e[i];i++)if(r.contentWindow===a.source){var d=a.data["datawrapper-height"][t]+"px";r.style.height=d}}});
Some sellers are recognizing the market has changed and others are not . . . I have one seller who overpaid for their home a few years ago and wants to list it at $950,000. The problem is recent comps call for a list price of $825,000,” writes Connie Durnal, a Redfin Premier real estate agent in Dallas. “I have another seller who paid $400,000 for their home but was willing to list it at $385,000, which was a great strategy. Because the home was fairly priced, it got multiple offers and sold for $10,000 over the asking price.
Redfins analysis is based on annual MLS data comparing original list prices with final sale prices. The firm didnt publish data for every metro.
These are tough times for many businesses across corporate America, many of whom are cutting down on business travel, and perks on the road. And in these times, one company’s policy on business travel is going viral: According to a recent Wall Street Journal article, Cracker Barrel employees reportedly must follow a new policy that they can only eat at Cracker Barrel restaurants while traveling for work.
But according to Cracker Barrel, that’s not exactly true.
“The policy for employees to dine at Cracker Barrel while traveling for business, whenever practical based on location and schedule, is not new,” Cracker Barrel explained to Fast Company in an email statement. “Also, it is not the only place that our employees may eat when on the road, as previously reported. The change was to further limit reimbursement of alcoholic beverages under the policy.”
Still, backlash to the reported policy comes during a rough patch for the American restaurant chain known for its Southern charm, marked by declining sales, and more customer backlash over a recent botched attempt to rebrand.
In August, Cracker Barrel unveiled a new campaign starring country music artist Jordan Davis that revamped its “Old Timer” logo and menus, and lightened up the restaurant’s dining rooms, to the dismay of longtime customers. (The reaction can be summed up by one TikTok user who posted, I prefer the darker cozier look, I also dont like change.”) The company was soon forced to walk back the plans, and later said it wouldn’t change the logo.
Cracker Barrel financials
Shares of Cracker Barrel (NASDAQ: CBRL) were down less than 1% in midday trading on Tuesday at the time of this writing.
The Tennessee-based chain’s first quarter fiscal 2026 earnings missed expectations, with total revenue at $797.2 million, down 5.7% compared to the prior year first quarter; same-store restaurant sales down 4.7% over the prior year quarter, and comparable store retail sales down 8.5%.
PayPal is replacing CEO Alex Chriss with Enrique Lores, saying that the pace of change and execution at the company has not met board expectations over the past two years.
Lores has served as a PayPal board member for almost five years and has been board Chair since July 2024. He’s also spent more than six years as president and CEO of HP Inc.
The payments industry is changing faster than ever, driven by new technologies, evolving regulations, an increasingly competitive landscape, and the rapid acceleration of AI that is reshaping commerce daily, Lores said in a statement on Tuesday. “PayPal sits at the center of this change, and I look forward to leading the team to accelerate the delivery of new innovations and to shape the future of digital payments and commerce.
PayPal’s board thanked Chriss for his contributions, including the role he played to monetize Venmo and grow the Buy Now Pay Later business.
Lores will take over as PayPal CEO on March 1. David Dorman will serve as independent chair, effective immediately.
PayPal’s Chief Financial and Operating Officer Jamie Miller will serve as interim CEO until Lores assumes the position.
PayPal also reported its fourth-quarter results on Tuesday. The technology platform and digital payments company posted an adjusted profit of $1.23 per share on revenue of $8.68 billion. The performance missed the expectations of analysts polled by Zacks Investment Research, who were looking for a profit of $1.29 per share on revenue of $8.77 billion.
The San Jose, California-based company also forecast lower profit for the first quarter.
Shares slid 16% before the market open.
Michelle Chapman, AP business writer
Amid nationwide outrage over the killings of Renée Good and Alex Pretti, two House Democrats are pressing Google and Meta to answer for recruitment campaign posts that Immigration and Customs Enforcement has recently run on their platforms. The lawmakers, Reps. Becca Balint of Vermont and Pramila Jayapal of Washington, have accused the companies of being complicit with the Trump administration and enabling ICEs efforts to promote slogans thatthey sayhave also been employed by white nationalist and neo-Nazi groups.
The inquiries were sent on January 21, and as of Monday, the platforms still had not responded. “What is going on with ICE is a five-alarm fire for our democracy, and these corporations are in it up to their necks,” Balint tells Fast Company. “They can no longer claim they ‘didn’t know.’ They are not only profiting from cruelty but actively helping to perpetuate it at everyone else’s expense. We expect answers, and we expect them now.”
Under the Trump administration, ICE has sought to rapidly scale up recruitment. The agency aimed to spend $100 million on the effort, according to a document reported by The Washington Post last year, and it outlined a wartime recruitment strategy that included targeting people who show interest in firearms, Ultimate Fighting Championship (UFC) events, and podcasts focused on patriotism.
ICE has run about 65 different advertisements on Google since the beginning of the year, according to the platforms ad library. These posts include a $50,000 signing bonus offer, opportunities to Defend the Homeland, and heavy use of Uncle Sam imagery. ICEwhich Rolling Stone reports has spent at least a few hundred thousand dollars running ads on Meta platforms in recent monthshas used its Facebook account to post provocative imagery alongside recruitment posts. These include posts featuring a picture of knights with swords alongside the text, THE ENEMIES ARE AT THE GATES,” as well as another displaying a man riding a horse and the phrase, WELL HAVE OUR HOME AGAIN.” Some of the posts are more explicit, including one showing a man carrying the Betsy Ross flag with the message, SEND THEM BACK.
The politicians’ letter to the companies aims to draw a direct line between Big Techs ad systems and the normalization of rhetoric that civil rights groups say echoes white supremacist propaganda. Just last week, DHS posted a recruitment ad on Instagram proclaiming well have our home again, which is a song popularized in neo-Nazi spaces and used in white nationalist calls for a race war. The same lyrics were found in the manifesto of Ryan Christopher Palmeter, the white supremacist who shot and killed three black people in Jacksonville in 2023, wrote Balint and Jayapal in their January letter to Meta. It appears Meta is complicit in furthering this content on behalf of the Trump administration.
These Facebook posts have racked up tens of thousands of likes or shares.
Though Google, which also owns YouTube, and Meta, which owns both Facebook and Instagram, are the platforms the lawmakers focused on, theyre not the only place where ICE has posted content. The agency has posted job ads or recruitment content on LinkedIn, which didnt respond to a request for comment. It’s not immediately clear that these platforms are the primary way the agency is actually finding new recruits. Still, the letter highlights that platforms stand to be drawn into the nationwide discussion over ICE and its tactics.
The companies confirmed receipt but havent responded yet, Balints office tells Fast Company. Meta declined Fast Companys request for comment, and Google did not respond to multiple requests for comment.
The silence isnt necessarily surprising. Tech companies have a real interest in not ruffling feathers with the Trump administration, and some platforms have, in the aftermath of the 2020 election, already done a major about-face about their decisions to boot or suppress the presidents account. Balint’s and Jayapals letter isnt a new strategy for lawmakers either. Members of both parties have previously pushed platforms to censor or restrain posts that they find odious. In highly polarized times, critics argue that this approach essentially amounts to working the refs, and it seems unlikely Google and Meta would move to censor an official government agency.
Ignaz Semmelweis was a physician working in a maternity ward in the 1840s. He noticed something disturbing: women giving birth in the ward staffed by doctors and medical students died from “childbed fever” at rates of 10-35%, while a nearby ward staffed by midwives had death rates under 4%.
The key difference was that doctors were coming straight from performing autopsies to delivering babies, without washing their hands. They would dissect cadavers in the morning, then examine pregnant women in the afternoon with just a quick rinse. In 1847, Semmelweis instituted a policy requiring doctors to wash their hands with a chlorine solution between the autopsy room and the maternity ward. Death rates plummeted dramatically to around 1-2%.
Great news, right? But instead of celebration, the medical community mocked Semmelweis for his claim that handwashing was worth the time and effort. He was driven out of the profession, and the childbed fever deaths went back up. It took more than 50 years after his discovery for handwashing to go mainstream in hospitals.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/speakeasy-desktop.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/speakeasy-mobile.png","eyebrow":"","headline":"\u003Cstrong\u003ESubscribe to Urbanism Speakeasy\u003C\/strong\u003E","dek":"Join Andy Boenau as he explores ideas that the infrastructure status quo would rather keep quiet. To learn more, visit \u003Ca href=\u0022http:\/\/urbanismspeakeasy.com\/\u0022\u003Eurbanismspeakeasy.com.\u003C\/a\u003E","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"http:\/\/urbanismspeakeasy.com\/","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91453933,"imageMobileId":91453932,"shareable":false,"slug":""}}
The case for cameras
Right now, in early 2026, state legislatures across the country are trying to outlaw a proven treatment for traffic injuries and fatalities.
Speed enforcement cameras are proven to reduce vehicle speeds and reduce crashes. According to the US Department of Transportations Proven Safety Countermeasures initiative, fixed speed cameras can cut crashes on urban principal arterials by up to 54% for all crashes and 47% for injury crashes.
For obvious reasons, school zones are the first place communities tend to install safety cameras. Speeding near schools creates unacceptable risks for kids crossing streets or waiting at bus stops.
Montgomery County, Marylands, automated speed enforcement program found that cameras reduced the likelihood of a crash involving a fatality or incapacitating injury by 19%, decreased the chance of drivers exceeding the limit by more than 10 mph by up to 59%, and fostered long-term changes in driver behavior that substantially lowered overall deaths and injuries.
In New York City school zones, fixed cameras have reduced speeding by up to 63% during active enforcement hours. Many other case studies demonstrate similar outcomes. The bottom line is automated speed enforcement saves lives.
Pre-installation surveys at some Virginia schools revealed a whopping 95% of drivers were blazing through school zones at 10+ mph during arrival and dismissal. Nearly every driver was risking the lives of young kids, including parents. In Fairfax County, the safety cameras at Key Middle School issued 7,429 citations from August 2024 to May 2025. But after the cameras had been in place for a while, average speeds fell from 33.1 mph to 27.8 mph.
People need consequences for dangerous driving. Automated cameras deliver fair, unbiased enforcement where officers can’t patrol constantly, holding reckless drivers accountable in high-risk areas like school zones while freeing up police for other duties.
Bills to ban
But while automated enforcement is saving lives, politicians in multiple states are advancing bills to ban, restrict, or phase out speed cameras.
Virginia: SB 297 (introduced January 13, 2026) repeals the authority for law-enforcement agencies to use photo speed monitoring devices. It has been referred to the Senate Committee on Transportation and remains under consideration in the 2026 Regular Session.
Arizona: SCR 1004 (advanced through the Senate Appropriations, Transportation, and Technology Committee in mid-January 2026) aims to place a statewide ban on photo radar enforcement (including speed cameras) on the November 2026 ballot for voter decision.
Georgia: HB 225 repeals all laws authorizing automated traffic enforcement safety devices (speed cameras) in school zones, with an effective date of July 1, 2028, to phase out existing contracts. Reintroduced in the 2025-2026 Regular Session (published January 13, 2026), it previously passed the House 129-37 in 2025 but stalled in the Senate.
Texas: Building on the state’s existing prohibitions on most fixed speed and red-light cameras (banned statewide in 2019), recent efforts like HB 2810 (introduced in the 2025 session but died) sought to expand bans to include portable devices enforcing speed limits. Similar measures could resurface in the 90th Legislature starting January 2027, driven by complaints about distractions from flashes and potential safety risks in local deployments.
Minnesota: Rep. Greg Davids (R-Preston) announced in late 2025 that he would author a bill to ban automated speed cameras statewide, to be introduced in the 2026 legislative session. This follows Rochester’s City Council narrowly approving a request for a speed camera pilot program, highlighting opposition amid concerns over enforcement fairness and local authority.
Robust evidence from federal and local sources supports speed cameras as effective for slowing drivers and preventing crashesespecially in child-heavy school zones. Its a shame to see politicians working to dismantle them.
Speed enforcement cameras save lives. The victims and survivors of traffic violence deserve better than the misguided bills that will directly lead to more life-altering crashes.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/speakeasy-desktop.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/speakeasy-mobile.png","eyebrow":"","headline":"\u003Cstrong\u003ESubscribe to Urbanism Speakeasy\u003C\/strong\u003E","dek":"Join Andy Boenau as he explores ideas that the infrastructure status quo would rather keep quiet. To learn more, visit \u003Ca href=\u0022http:\/\/urbanismspeakeasy.com\/\u0022\u003Eurbanismspeakeasy.com.\u003C\/a\u003E","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"http:\/\/urbanismspeakeasy.com\/","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91453933,"imageMobileId":91453932,"shareable":false,"slug":""}}
Firefox has a reputation as the browser of choice for power users who prefer to customize everything and it just gave users one very important new option. While most other tech companies shove AI enhancements down their users throats, Mozilla is introducing a way to disable Firefoxs AI features outright a boon for anyone searching for a safe haven from the AI software onslaught.
Starting on February 24 with the Firefox 148 update, users will be able to toggle AI off in a new AI controls area in the desktop browsers settings menu. To disable AI, you wont even need to dig around and disable features one by one: Mozilla describes the forthcoming option as a single place to block current and future generative AI features across Firefox.
If youd like to customize Firefoxs AI offerings, the browser will also allow you to check and enable individual features. In a blog post announcing the option, Mozilla recognizes that not everyone wants to use AI, but it will continue to work on AI features for Firefox users who do want them. The options on the way later this month will allow Firefox users to toggle AI on or off for translation tools, alt text descriptions in PDFs, tab groups, link preview summaries, and for a sidebar feature that incorporates chatbots like ChatGPT, Claude, and Gemini.
Mozilla has been tinkering with AI product experiments in Firefox for a bit now. The company began rolling out access to AI chatbots a year ago with Firefox 135 and last September invited iOS users to shake to summarize a website with AI.
Mozilla walks a tightrope on AI
Mozilla announced its plan to splice AI features more deeply into Firefox late last year, a decision panned by some of its users. At the time, the company emphasized that any AI tools would be opt-in and designed to keep users in full control. … We believe AI should be built like the internet open, accessible, and driven by choice so that users and the developers helping to build it can use it as they wish, help shape it and truly benefit from it, Mozilla wrote in the announcement.
The Firefox maker just appointed a new CEO as the company promotes its image as the worlds most trusted software company. Mozilla tapped Anthony Enzor-DeMeo, previously the general manager of Firefox, to step into the role. Enzor-DeMeo described the browser as the next battleground for AI in a statement paired with the news. Its where people live their online lives and where the next eras questions of trust, data use, and transparency will be decided.
Firefox users are paying close attention. Mozillas connections to the open source community and its emphasis on user choice have built a deep well of brand loyalty over the years. Still, AI is a divisive technology, and one that Firefox users arent all sold on a fact the browser maker is well aware of.
We believe choice is more important than ever as AI becomes a part of peoples browsing experiences, Head of Firefox Ajit Varma wrote in Mozillas announcement on AI controls. What matters to us is giving people control, no matter how they feel about AI.
Lior Pozin had an epiphany about AI infrastructure in early 2025. As CEO of AutoDS, an AI-powered e-commerce automation platform, he had pushed his team to deploy AI features quickly, betting that speed would define success.
AutoDS was bootstrapped and eventually reached 1.8 million users, generated more than $1 billion in user revenue, and exited successfully to Fiverr. From its earliest days, the company was fast moving, the kind of place where speed was strategic and rapid implementation felt like the natural way to operate. But as Pozins team moved from pilots to production, they learned that speed alone was not enough. AI only delivers results when the right data foundations and ownership structures are in place.
Without the right governance, data organization, and access, AI cant scale, Pozin tells Fast Company. Once we built that foundation, everything changed. AI stopped being a feature and became part of how we operate.
That experience was not unique to AutoDS. In 2025, across several industries, companies quickly realized that deploying AI at scale required confronting uncomfortable truths about their infrastructure, their assumptions about what AI could do, and their willingness to solve unglamorous problems before chasing transformative ones. While the year began with big promises, it turned out to be less about breakthroughs and more about a reckoning with reality.
The lessons that emerged reveal an industry growing up. Instead of building ever more powerful models or simply raising more capital, the industry is maturing by figuring out what actually works when the demos end and the real work begins.
INFRASTRUCTURE FIRST, OR NOTHING ELSE MATTERS
In early 2024, database company RavenDB explored building an AI assistant for its documentation in collaboration with Microsoft. The project ultimately fell apart. According to founder Oren Eini, the problem was not the AI model itself but everything surrounding it.
Data had to move through multiple systems before reaching the model, and updates required manual intervention. The entire setup depended on fragile connections that could break at any moment. For a database company, the irony was hard to miss.
The experience clarified something essential for the team: AI needed to be integrated far more deeply into the database itself to be reliable, predictable, and scalable.
For Eini, it wasnt a setback so much as a signal that the surrounding architecture mattered as much as the model itself. That realization informed RavenDBs more recent work on AI agents in and capabilities built directly into the database layer, where models operate closer to the data they rely on and can behave more predictably in production environments.
At AutoDS, that shift translated into a more deliberate approach. The team focused on building a shared data layer into its drop-shipping platform and clearer ownership around AI initiatives, which later enabled products like its AI-powered store builder to scale more reliably across the business.
The shift required patience. Pozins team stopped chasing what looked impressive and started tracking what mattered: time saved, accuracy improved, and decisions accelerated. Success now means AI actually improves how we work, not just that were using it, Pozin notes.
EFFICIENCY BEATS RAW POWER
While much of the AI industry chased larger models and more compute in 2025, Oculeus, a software-for-telecom company with deep experience in AI, spent the year prioritizing efficiency. The team focused on designing and refining systems that deliver reliable performance without excessive computational overhead. That focus is central to how Oculeus applies AI in telecommunications, where its systems are used to detect fraud patterns and anomalous behavior in real time.
In those environments, Arnd Baranowski, the companys CEO, explains that predictability matters more than novelty, because false positives and inconsistent outputs carry direct financial and operational risk.
AI algorithms and technology, which go along with massive computation and energy consumption, are a misguided path, Baranowski adds. His critique extends beyond hardware, questioning the industrys embrace of nondeterministic systems that produce different outputs for the same input. Training must result in 100% deterministic responses. Otherwise, something is wrong.
That stance runs counter to the excitement around large language models, which treat randomness as a feature. For Baranowski, the lesson of 2025 was simple: AI systems only earn trust when they behave consistently and can be relied on in real operating conditions.
Eini also shares that view. At RavenDB, the goal wasnt building the smartest AI. It was building predictable AI that could handle routine tasks without drama. We dont necessarily want smart AI, Eini says. We want predictable AI.
As compute costs remain high and energy consumption becomes a public concern, 2026 will favor companies that figured out how to do more with less over those still chasing the biggest possible models.
TRUST DEMANDS BOUNDARIES
In 2024, Air Canadas chatbot promised a customer a bereavement fare discount that didnt exist. The airline was held liable. The case crystallized a problem that became unavoidable in 2025: AI agents cant be trusted the way employees can.
Eini frames it bluntly. A bank teller is bound by policies and consequences. An AI agent isnt. I like to think about them as employees who I know are susceptible to bribes, he says. Its crucial to consciously set boundaries for their actions and actively implement protective measures.
Those boundaries took practical form. At AutoDS, Pozin created a dedicated team to verify AI outputs and ensure the system received accurate source data. At RavenDB, the team developed and implemented chain-of-approval processes and clear limits on what AI agents could access or promise.
The lesson extends beyond technical safeguards. AI agents exist in a gray zone between tool and actor. They respond to instructions but lack judgment. They execute tasks, but cant weigh the consequences. That reality requires new frameworks for accountability that dont assume good training guarantees good behavior.
Organizations thriving in 2026 will treat AI deployment as a trust problem first. That means transparency about capabilities and limits, clear expectations for users, and systems designed to fail safely when things go wrong.
SMALL FIXES BEAT MOONSHOTS
The years biggest AI narratives centered on autonomous vehicles, artificial general intelligence (AGI)which AI scientist Yann LeCun thinks is an illusionand models replacing entire professions. But companies making actual progress focused elsewhere: solving small, annoying problems at scale.
The biggest changes will come from fixing many small problems, not from one big, all-knowing AI, Eini says. Quantity has a quality of its own, and removing many small frictions leads to a much faster pace overall.
RavenDB empowered regular team members to build AI features in days rather than waiting for top engineers to approve and execute. AutoDS measured success by whether AI made employees faster and more efficient, not by how many AI projects were running. The results were individually modest but collectively transformative.
A year earlier, companies chased AI for its own sake, deploying pilots that looked impressive in demos but never scaled. In 2025, the focus shifted to measurable impact. Eini compares it to how we today make water potable for drinking, a practice so ordinary now that no one thinks about it. In the same sense that ATMs or self-checkout services havent fundamentally changed the entire world, but have made our lives measurably better, I think well see a lot of that, he tells me. The sheer quantity of changes will have a transformative effect.
PREPARATION MATTERS MORE THAN REACTION
Steve Brierley wasnt building AI in 2025. As CEO of quantum computing company Riverlane, he was watching how unprepared industries were when ChatGPT arrived. The AI boom exposed how unready many industries were when tools like ChatGPT suddenly entered the mainstream, forcing companies to scramble around regulation, scalability, data readiness, and consolidation, and a widening workforce and skills gap, Brierley says.
His takeaway: understand emerging technologies early enough to anticipate challenges rather than react to crises. Quantum computing will arrive sooner than many expect, and it wont be a marginal improvement. AI excels at analyzing and generating insights from data, while quantum computing will enable the creation of new kinds of data altogether, Brierley says. Together, they will unlock far greater exploration, discovery, and innovation than technology could achieve on its own.
Gilles Thonet, deputy secretary-general at the International Electrotechnical Commission, saw the same dynamic in regulation. As AI laws took effect in 2025, companies struggled to translate legal requirements into operational reality. International standards are essential to fostering trust in this transformative technology, Thonet says.
WHAT COMES NEXT
The lessons from 2025 point toward an AI future grounded in operational reality rather than hype. Companies leading that shift built infrastructure, set boundaries, and solved real problems instead of chasing headlines.
But new challenges are emerging. Sheetal Mehta, global head of cybersecurity services at NTT Data, warns that AI capabilities driving productivity gains are being weaponized. Agentic AIs speed and ability to learn and make decisions autonomously can also be used by cybercriminals, exposing enterprises to new attack surfaces and unexpected security vulnerabilities, Mehta says.
That means 2026 will require better safeguards, not just better systems. Organizations will need to treat AI security, governance, and ethics as foundational, not optional.
Pozin captures that shift rather poignantly. The next phase of AI is AI that lives with us, learns us daily, and delivers exactly what we need, just in time. It wont feel like a tool anymore. Itll feel like a teammate that truly gets you, he says.
Eini puts it even more simply: Moving beyond the initial awe to become a transparent tool that simply gets things done.
Not AGI. Not full automation. Just AI that works reliably, scales predictably, and solves problems without creating new ones. For an industry that spent years chasing moonshots, that might be the most ambitious goal of all.
In France, civil servants will ditch Zoom and Teams for a homegrown video conference system. Soldiers in Austria are using open source office software to write reports after the military dropped Microsoft Office. Bureaucrats in a German state have also turned to free software for their administrative work.
Around Europe, governments and institutions are seeking to reduce their use of digital services from U.S. Big Tech companies and turning to domestic or free alternatives. The push for digital sovereignty is gaining attention as the Trump administration strikes an increasingly belligerent posture toward the continent, highlighted by recent tensions over Greenland that intensified fears that Silicon Valley giants could be compelled to cut off access.
Concerns about data privacy and worries that Europe is not doing enough to keep up with the United States and Chinese tech leadership are also fueling the drive.
The French government referenced some of these concerns when it announced last week that 2.5 million civil servants would stop using video conference tools from U.S. providers including Zoom, Microsoft Teams, Webex, and GoTo Meeting by 2027 and switch to Visio, a homegrown service.
The objective is to put an end to the use of non-European solutions, to guarantee the security and confidentiality of public electronic communications by relying on a powerful and sovereign tool, the announcement said.
We cannot risk having our scientific exchanges, our sensitive data, and our strategic innovations exposed to non-European actors, David Amiel, a civil service minister, said in a press release.
Microsoft said it continues to partner closely with the government in France and respect the importance of security, privacy, and digital trust for public institutions.
The company said it is focused on providing customers with greater choice, stronger data protection, and resilient cloud services ensuring data stays in Europe, under European law, with robust security and privacy protections.
Zoom, Webex and GoTo Meeting did not respond to requests for comment.
French President Emmanuel Macron has been pushing digital sovereignty for years. But theres now a lot more political momentum behind this idea now that we need to de-risk from U.S. tech, Nick Reiners, senior geotechnology analyst at the Eurasia Group.
It feels kind of like theres a real zeitgeist shift, Reiners said
It was a hot topic at the World Economic Forum’s annual meeting of global political and business elites last month in Davos, Switzerland. The European Commission’s official for tech sovereignty, Henna Virkkunen, told an audience that Europe’s reliance on others can be weaponized against us.
Thats why its so important that we are not dependent on one country or one company when it comes to very critical fields of our economy or society, she said, without naming countries or companies.
A decisive moment came last year when the Trump administration sanctioned the International Criminal Court’s top prosecutor after the tribunal, based in The Hague, Netherlands, issued an arrest warrant for Israeli Prime Minister Benjamin Netanyahu, an ally of President Donald Trump.
The sanctions led Microsoft to cancel Khan’s ICC email, a move that was first reported by The Associated Press and sparked fears of a kill switch that Big Tech companies can use to turn off service at will.
Microsoft maintains it kept in touch with the ICC throughout the process that resulted in the disconnection of its sanctioned official from Microsoft services. At no point did Microsoft cease or suspend its services to the ICC.
Microsoft President Brad Smith has repeatedly sought to strengthen trans-Atlantic ties, the company’s press office said, and pointed to an interview he did last month with CNN in Davos in which he said that jobs, trade and investment. as well as security, would be affected by a rift over Greenland.
Europe is the American tech sectors biggest market after the United States itself. It all depends on trust. Trust requires dialogue, Smith said.
Other incidents have added to the movement. There’s a growing sense that repeated EU efforts to rein in tech giants such as Google with blockbuster antitrust fines and sweeping digital rule books haven’t done much to curb their dominance.
Billionaire Elon Musk is also a factor. Officials worry about relying on his Starlink satellite internet system for communications in Ukraine.
Washington and Brussels wrangled for years over data transfer agreements, triggered by former National Security Agency contractor Edward Snowdens revelations of U.S. cyber-snooping.
With online services now mainly hosted in the cloud through data centers, Europeans fear that their data is vulnerable.
U.S. cloud providers have responded by setting up so-called sovereign cloud operations, with data centers located in European countries, owned by European entities and with physical and remote access only for staff who are European Union residents.
The idea is that only Europeans can take decisions so that they cant be coerced by the U.S., Reiners said.
The German state of Schleswig-Holstein last year migrated 44,000 employee inboxes from Microsoft to an open source email program. It also switched from Microsoft’s SharePoint file sharing system to Nextcloud, an open source platform, and is even considering replacing Windows with Linux and telephones and videoconferencing with open source systems.
We want to become independent of large tech companies and ensure digital sovereignty, Digitalization Minister Dirk Schrödter said in an October announcement.
The French city of Lyon said last year that it’s deploying free office software to replace Microsoft. Denmarks government and the cities of Copenhagen and Aarhus have also been trying out open-source software.
We must never make ourselves so dependent on so few that we can no longer act freely, Digital Minister Caroline Stage Olsen wrote on LinkedIn last year. Too much public digital infrastructure is currently tied up with very few foreign suppliers.
The Austrian military said it has also switched to LibreOffice, a software package with word processor, spreadsheet and presentation programs that mirrors Microsoft 365’s Word, Excel and PowerPoint.
The Document Foundation, a nonprofit based in Germany that’s behind LibreOffice, said the military’s switch reflects a growing demand for independence from single vendors. Reports also said the military was concerned that Microsoft was moving file storage online to the cloud the standard version of LibreOffice is not cloud-based.
Some Italian cities and regions adopted the software years ago, said Italo Vignoli, a spokesman for he Document Foundation. Back then, the appeal was not needing to pay for software licenses. Now, it’s the main reason is to avoid being locked into a proprietary system.
At first, it was: we will save money and by the way, we will get freedom, Vignoli said. Today it is: we will be free and by the way, we will also save some money.
Kelvin Chan AP business writer
Associated Press writer Molly Quell contributed to this report.
