The tech industry is in flux regardless of how the AI bubble resolves, and that instability now shapes the daily working lives of almost everyone who builds for the web. This article is about navigating that situation — but first, a clear-eyed look at where we actually are.
What changed, and what's still being asked of us
Hyperscalers like Microsoft and Meta, alongside AI startups like Open AI and Anthropic, have transformed everything about tech and the work we do — whether or not they succeed in getting everyone to pay a subscription for their Large Language Model (LLM) technology, which is marketed to us as AI. For simplicity, this article calls it AI throughout. The effects reach well beyond our industry: LLM-generated slop shows up on flyers, websites, village posters and food truck menus, and people are even using LLMs to work out what's special about their own children.
In tech, the impact has been especially hard. Mid-last year I investigated because I suspected AI mandates were spreading; what I learned in those conversations feels like utopia compared to now. In the time since that piece was published, the discourse moved from token maxxing to companies realizing the technology is extremely expensive and not returning the promised outcomes. The mandates have cooled, but demands to use AI to produce more volume — needed or not — remain. Friends across the industry, from hyperscalers to tiny studios, describe work lives that are nothing short of horrifying, and few people seem happy with their work. That's a long way from even a handful of years ago, when the culture and outcomes were often problematic but people were generally happy doing the work.
The cost to our communities
The sprint toward AI has had devastating effects in many places: it has encouraged suicide, devastated communities with the data center build-out and displaced thousands of careers. Those are only a tiny sample of the harms; listing them all would make this article unbearably long and depressing.
The damage to our communities in tech matters just as much. Open source and open knowledge — people sharing knowledge and building everyone up with them — was the foundation of the industry's, especially the web industry's, rapid success. Those people still exist, myself included, but being one of them is increasingly difficult. As the numbers go up in the AI world, they go very much down for independent educators, publishers and community conferences. There is also a worrying culture of "I don't need to skill up now because I can use AI for work" — a reductive view of your work and career that is harmful in the long term.
We are more than code slingers
Typing characters into a text editor is not what makes you a good developer. Experience, knowledge, nuance and above all creativity do. AI can feel like a miracle machine, but it's important to take a wider view. Alongside the published studies on cognitive decline and the widening skills gap, investing in your brain matters more than ever: if AI tools do everything for you, you no longer get the repetitive reinforcement that writing code every day provides. That's why I build most of my teaching around a linear build-along — repeating patterns you've learned makes the knowledge sink in.
To founders, decision makers and managers specifically: the most important part of our job is ensuring our workers are looked after. Not just compensation and perks, but keeping their skills sharp, satisfying their professional curiosity and, above all, ensuring they are happy in the work. Our workers are much more than "number go up" vehicles. Give developers more time to learn instead of chasing endless volume. If AI enthusiasts are to be believed, the technology creates hours of time efficiencies — don't refill that time with more volume. Invest it in your people's skills. It will help avoid the burnout that daily LLM use will likely cause in the long term.
Why sprinting toward LLMs is risky
The health of the AI industry is not good. The bubble has already effectively burst in China, and more worryingly, the "picks and shovels" — the chips industry in Korea — has seen dramatic downturns. That is not how an "inevitable" outcome is supposed to happen.
The U.S. AI industry sits on very shaky ground too: circular financing, unbelievable levels of debt and a data center build-out that cannot keep up with the power levels required to turn profits for hyperscalers and startups, mixed with genuine public hatred of the technology. Recent events — Democrats dominating the midterms and AI stocks dropping, as well as Nvidia's stock falling because its CEO was caught up in a spying scandal — highlight the industry's fragility. A healthy industry doesn't work like this. So what would be a really silly thing to do? Put AI at the heart of everything in your business. Open source models are, I think, the logical destination of the technology, but few companies will be willing to maintain their infrastructure needs; the low barrier of entry of commercial AI is its real selling point.
Investing in people instead of replacing them
The starting point for any response is that people must be invested in, rather than set aside by technology such as AI. Experience, knowledge, nuance and creativity are what separate humans from AI, and while the technology keeps evolving, even its strongest advocates accept that the rate of frontier model improvement is leveling out considerably. Human skills therefore need to stay sharp, which suggests several concrete actions.
Learning as a funded activity
The responsibility here falls on founders, decision makers and managers as much as on individuals. The industry once had a strong record of investing in skilling people up, and that needs to return. Staff should get both the time and the budget to learn, develop their skills and engage with the community at conferences. The benefits are already well understood; what matters is bringing the practice back into the organization. At worst, it makes AI usage inside a company more efficient — and more cost effective once token subsidies inevitably end.
Revisiting AI mandates
Those subsidies will not last. Getting thousands of dollars of value from a $200-a-month subscription will become a fond memory, and once hallucinations from AI tools cost thousands, pushing staff to produce maximum volume no longer makes sense. Past decisions should be re-evaluated, with saved costs invested back into developing workers while there is still hope of avoiding complete burnout. Workers belong in those conversations: the impact of mandates on them needs to be discussed and understood.
Doing the work by hand
For developers especially, repetition — typing above all — reinforces both new learning and existing knowledge. The slower pace also gives the brain room to think, which matters for output quality; skipping all of it atrophies skills over time. Jenn Schiffer's account of the Weird Web October challenge makes the case for practice without the tools:
i encourage anyone who uses ai and/or vibe codes to consider doing an exercise like this without those tools as a way to challenge your brain and you can continue doing your regular work the way you do your regular work
Even a small side project works, assuming there is energy left after producing value all day. For managers, it is another way to make staff happier: give them time to do this on the clock.
Funding the independent side of the industry
The soul of the industry will be missed when it is gone, but it is far from gone yet. Independent outfits such as CSS-Tricks, Piccalilli and Smashing Magazine work hard to keep it alive by sharing high-quality educational content. Most of that content is free at the point of entry but not free to produce: writers are paid fairly, the organizations are tiny, and nobody is in it for large profits. Funding is the ultimate barrier to doing more.
Companies and individuals can help by purchasing courses, joining membership programs and buying ad space. With the right financial support, small publications can become a real force for pushing good practice back into the industry and bringing everyone up with it.
Course producers worth considering include Miriam, Sara, Mat, Scott, Brad and Ian, Josh, Ahmad, Christine, Vitaly, Ben and Andy Clarke himself. Many will run generous Black Friday deals — a good moment to ask the boss to invest in you.



