GarageAI.tl;dr: Deepseek & AIs first Next-Cycle narrative I’m sending Off Kilter a bit early this week because we’ve just had a seismic leveling of the playing field in AI. Normal scheduling shall return next edition. By now, you’ll almost certainly have heard of Deepseek, the Chinese corporation that just released a GenAI model that’s equivalent to the best of the West, but is 40X more efficient. The big story isn’t the drop in Big Tech market value, nor is it the seismic shift in the geopolitics of AI. Nope. The big story is that now we know cheap models are possible, it’ll be replicated and improved upon, creative destruction will happen, and the narrative around generative AI will rapidly shift from Big Tech dominance to the competitive commoditization of a core technology of the future. Don’t believe the inevitable spin that’s coming. This is only an existential crisis for Big Tech. Everyone else will pivot because the commoditization of core technologies benefits everyone but the very richest: while the process will be messy, it will drop costs, drive innovation in specialist LLMs, increase competition, and unleash the first Next-Cycle startup narrative of the AI era. I’m calling this phenomenon ‘GarageAI,’ because it won’t be constrained, directed, nor controlled by the demands of the behemoths. To explain GarageAI, let’s first walk back through history. If we zoom out toward the long arc of its history, creative destruction has been the OS that drove Silicon Valley. New technologies replaced the old, spawning a new breed of innovators who’d turn a garage startup into a multi-billion dollar corporation, growing while the those created by the previous generation of innovators slowed, declined, or died. It spawned stirring tales of the American dream, derring-do, and founder-innovators as pop culture heroes. Then, a very small number of the current generation became so powerful that they gained monopolistic dominance across a vast swathe of the global economy. Acutely aware that creative destruction is the OS of Silicon Valley because they benefited from it, the leaders of these firms then spent years ensuring that no garage startup would ever happen to them. This is important because when OpenAI began to have breakthroughs with its eye-wateringly expensive blunt-force models, the Big Tech response wasn’t to view this as something to be improved upon but to treat it as a feature to be doubled and tripled down on, for it meant they’d be the one’s picking the winners because the next wave of technological disruption would be closed to all but the largest and richest corporations on earth. In other words, them. The sheer scale of their Last-Cycle ambition, hubris, and propaganda served its purpose. It chilled Western entrepreneurialism at the model level, pushing AI startups toward either infrastructure, primarily chips, or new application products that wrap around the core models instead. This works for Big Tech because rather than worrying about a new generation of entrepreneur replacing them, their riches would give them control over infrastructure and make anything built atop their models feudally dependent. And from there, we know the playbook: Let the startups be entrepreneurially dynamic, then hit them with ever-increasing prices and terms of service changes once they begin sniffing success because they’d have no other option but to pay—not enough to put them out of business, but enough to put them on life support. All this would be justified by the vast cost of the AI models, which would, of course, have to be recouped. Copy and featurize the startups that look like they might be successful in cracking the scale markets Big Tech requires for its own growth. Buy and take off the board any startups that look like they might threaten Big Tech’s existing franchises.
Then, the Chinese figuratively ‘blew the doors off’ by developing an entirely new approach that didn’t require the cost and power that Big Tech thought was necessary, ironically, because of the creative constraint of not being able to access the most powerful chips because of US export controls extensively lobbied for by…Big Tech. Oh my. Make no mistake, this changes pretty much everything. While the media narrative will likely be about US/China relations, and will almost certainly make the mistake of confusing capital for innovation, the real story is that Silicon Valley is about to have an immune response to big tech monopolization as its core DNA of creative destruction kicks in with a vengeance and GarageAI flourishes. It’s not just about the models; it’s about competition making these models predictably cheap enough to radically change the economics, and thus the entrepreneurial velocity, of the coming productization and commercialization of GenAI into everyday life. This means GenAI entrepreneurs will now be able to embrace its Next-Cycle narrative free from dependence on the Last-Cycle Big Tech firms they may one day displace; the question now is what it's going to take for such startups to break through in a sea of new competition that’s about to emerge. It’ll take more than just a great product, as this is increasingly the cost of entry in tech. So, if any of you are out there, here are a few things to think about: Do not ape big tech. Big tech displays all the negative hallmarks monopolists have displayed since the British East India Company in 1600: No longer innovators but anti-innovators. No longer agents of progress but agents of anti-progress. While you may think that copying it will replicate its success, it will not, and for one simple reason: It only gets away with the behavior it exhibits because of monopoly power, and the defining characteristic of all monopolies is that they’re fundamentally anti-human. You will not beat an anti-human Big Tech monopolist by you, too, being anti-human, for you do not have its power. Instead, you will win by being its opposite, its alternative. Where it seeks to force, you must seduce. Thus, the GarageAI firm of the future must be relentlessly pro-human in all facets, from its management style to its product experience and approach to market, if it is to succeed. Beware the SaaS trap. These corporations ignored a hundred years worth of marketing knowledge and piled billions of dollars worth of ZIRP-era VC funding into big tech marketing and advertising products, believing it represented a new frontier of efficient customer acquisition. Now, they’re mostly on life support, caught between two equally unappealing truths: The transactional, euphemistically labeled ‘performance marketing’ techniques they’re optimized for aren’t working: Costs are up, response rates are down, ROI has become a sad tale of diminished returns. Yet, they cannot change course because they’ve become dependent upon tactics that no longer drive growth, and if they turn them off, they lose what they already have. Worse, through short-termism and widespread ‘flood the zone with shit’ marketing tactics, they turned the open web they relied upon for business into a pit of mediocrity, accidentally turning people away…and straight into the hands of AI. Build your brands from day one. Align your Big-B and little-b brand, and craft a strategy explicitly designed to ensure that you’re in charge of your own destiny rather than becoming dependent upon the monopolists you seek to disrupt. This is the true lesson from AirBnB shifting most of its marketing efforts from ‘underperformance marketing’ to brand marketing. While its biggest competitor, Booking.com, acts like a drug addict, dependent on Google AdWords to the tune of billions a year, over 90% of AirBnB traffic is organic. It controls its own destiny, Booking does not. Here’s how Big Tech got to businesses like Booking. It persuaded an entire generation of business leader that its advertising products were so efficient they didn’t really need much of a brand. Don’t get sucked in. If you squint at it in the shortest of short terms, at a skirmish level, they may have a point, but zoom out to the longer haul across which competitive battles are truly won and lost, and brand strength rips these efficiency claims into tiny little pieces. Be special. There’s a very simple takeaway from the empirical work of the marketing science community: Large businesses gain market benefits solely from their scale. Small businesses gain market benefits by being special, and it’s this specialness that enables growth. If you’re large and special, it’s extra good, but specialness is hard to maintain at scale. Here’s the kicker. If you’re small and not special, you get no benefit at all. You die. Unfortunately, when I look at emerging AI companies, I see a sea of same. Hiding your specialness behind your pitch to investors will do nothing to help you win with people and everything to make you lose. You must stand out to break through. You must be uniquely you, not some facsimile of others. And you must be seductive, for why else would anyone invite you to a party where Big Tech is already there trying to down all the booze and is acting like a dick? This is a generational-level disruption, freeing AI from the shackles of the Last-Cycle and pushing it into the arms of the Next, as Silicon Valley's garage-driven DNA belatedly kicks in with a creatively destructive immune response to monopoly. I can’t wait. Let’s go.
|
