Back in August, Noah published his 2026 update to “things I think I think about AI” and it was a certified banger that sparked lively discussion within our team. To capture that spirit in this follow-up piece, I gave the team three guidelines to share their hottest takes about AI.
1) Don’t think too hard about it – I would love for these to be more off-the-cuff opinions that live rent-free in your head
2) Don’t worry about backing up the claims too much with anything but personal experience– these are quick, pithy, vibe based takes.
3) The more random, the better! I’d love for this to feel a little bit frenetic and incorporate a diverse range of perspectives.
What follows is 25 things we think we think about AI across the broader Alephic/BRXND family.
But first BRXND NYC is now officially three weeks out and we have limited tickets remaining. Thanks again to Runway, Emberos, VantagePoint, Dstillery and WRITER for supporting the show. Can’t wait to see many of you there- we will have a sold out show so grab a ticket today before we have to move back to a waitlist.
On AI & the future of work
As the models, harnesses, and general UX of AI improves, the alpha that comes from mastering the more idiosyncratic aspects of working with AI (prompting, skills, model routing, etc.) will trend toward zero. Essentially, everything the AI guru-industrial complex espouses will become less valuable. The alpha of being generally AI-pilled, but not dogmatic about any one tool and able to imbue AI work with original, contrarian thought, will only increase. — Mike
AI is very good at continuing past the point where it should stop. It will happily add another abstraction, another paragraph, another ticket, another “improvement.” Knowing what not to build, not to say, or not to preserve is becoming more valuable, not less. — Claire
We’re approaching a point where we can accomplish more in a day than we can cognitively absorb. I’m not sure what happens to our experience of work when we can no longer fully process what we’ve accomplished. — Joh
Burnout is going to multiply significantly in the AI era. As AI automates more of the redundant tasks that used to serve as a brainless mental reset, we’re spending more and more cognitive energy every day. This is unsustainable in ways that aren’t obvious yet. As a result, shorter, more intense stints and “reset breaks” will become far more common and normalized, first in tech careers and then more broadly. That will challenge a lot of our current labor infrastructure. — Mike
Not all friction is bad. Sometimes the work you’re trying to automate away is what makes you better at your job. Other times, the process itself is part of the fun, like digging through records instead of getting an instant curated playlist. Whether removing friction helps depends on what you’re trying to get out of it.— Joh
On building with AI
Coding agents are good at so much more than web apps. Think hard about where you can put the working part of what you’re trying to create: CLI, server, plugins, Apple Watch, phones, other agents. — Aaron
SaaS (or “managed applications/APIs”) as we know it today is eventually going away. It will transition to “headless” interfaces that LLMs integrate into applications they build on the fly. The transition will take longer in enterprise and corporate environments, but it’s already starting at the consumer level. — Ian
The fact that so many more people can build software now magnifies the tension between individual ownership of outcomes and collective ownership of systems. Within an organization, how do we decide what should scale, who owns it when others start depending on it, and how to introduce necessary rigor without slowing down the people building? I don’t think we’ve figured out how to reconcile those things yet. — Joh
Time to market is more compressed than ever, but I think people’s patience to find “the good one” will increase in proportion to the number of apps, products, and things being shoveled out there. — Aaron
We don’t talk enough about how good the video models are getting. I think in early 2027 we’ll see feature-length, fully AI-generated movies break out into mainstream success, not exactly because they’re AI, but because they’re just good. I’m really excited about lowering the bar to creative production— Tom
On the vibes around AI
The last 10 years of “AI-powered” marketing BS has ruined the definition and perception of AI for the general public. — Charles
Social media is doing AI a great disservice. My TikTok feed is filled with anthropomorphic fruit in Pixar-style melodramas: a lemon gives birth to an orange and the family erupts, accusing the mother of cheating; a banana fakes being rich and gets found out. Social media is a machine for surfacing whatever makes you say “ew” out loud, and it turns out generative video is exceptionally good at manufacturing that on demand, in volume, for free. So this is the AI most people see. They’re not being unreasonable when they conclude AI is worthless and a little vile; they’re reasoning correctly from the samples they’ve been given. — Claire
The paid layer doesn’t help either. Companies like OpenAI and Anthropic sponsor creators to talk about AI, people notice, and the reaction is “this is propaganda.” Both are true at once: it is paid placement, and paid placement is simply the native condition of social media, the same as every piece of #sponcon that came before it. But because AI is the subject, undisclosed sponsorship reads as conspiracy rather than business as usual. So the organic AI content is slop, the sponsored AI content is suspect, and there’s no third category that most people ever encounter. — Claire
On being human in the age of AI
AI is still not funny and has never made me laugh. This is notable: it’s one of the few areas that hasn’t budged while everything else has. Have a little whimsy! — Claire
The arts will never be the same, and live performance will grow in significance; displaying true human expression in meatspace will be increasingly valued. Musicians who consider themselves “just recording artists” need to buck up and play live if they want to survive (unless they’re already famous, or just supermodels). The value of a refined piece of recorded audio is dropping. — Charles
The world needs a new solution for proving that you’re a person. — Charles
On winners and losers in the AI era
Cloudflare and Stripe will be the two most important companies of this era for CMOs, outside of Google, Meta, and OpenAI. — Mike
Web analytics is dead, but no one has acknowledged it yet. There are two options: either we completely abandon analytics as a concept, or we standardize agent and bot sessions with a consistent identifier so we can see their browsing and searching behavior separately from real users. (What even IS a real user in 2026, anyway?) — Tom
Apple will be the largest winner of the AI boom and the first company to reach a $10T valuation.— Mike
The conversation around publishers, fair use, and content licensing is going to absolutely explode. As users rely on frontier models for agentic workflows, those models need to be able to access real content. This is a ticking time bomb. We’re going to see weird things like ChatGPT having access to the NYT and Anthropic not, or vice versa. It’s the AI-era equivalent of not being able to pay for things inside an app because of App Store and Play Store rules: it makes zero sense to the user and is a direct consequence of licensing negotiations. Or… we’ll see a robust content licensing ecosystem emerge. But I’d put the chances of that working at less than 5% at this point. — Tom
A few more deep cuts
The models are getting very good at writing. All the writers saying they’ll “never use AI in their workflows” are wrong. Just like Microsoft Word, spellcheck, and Google Docs, AI is going to be an essential, wonderful tool for augmenting writing. (But/and: we will continue to see slop as far as the eye can see.) — Tom
I think we’re barely scratching the surface on memory. You can fit most people’s useful context into a small .md file (most people are not that unique). Just like we talk about a “coding harness,” I think we’ll see innovation in a “memory harness” that auto-summarizes, compacts, extracts, and remembers. Importantly, this will push into new territory like values: beyond specific facts, what does the model know about what you value? And forgetting: how does it discard things you’ve outgrown? We need an overnight memory process. Do LLMs dream of electric sheep? — Tom
Gen Z and Gen Alpha aren’t going to experience some existential intellectual crisis because they were brought up AI-native in the workplace. The kids will still be able to think. Older generations complaining that the generation below them doesn’t have that dog in them is just a tale as old as time. — Mike
I sadly disagree with Mike. As a humble Gen Zedder myself, I have a lot of concern for my own peers, many of whom have decided to abstain from using AI merely because they can, and because it sounds cool to say so. Right now refusing costs nothing, which is exactly why it reads as trendy. I’ve met with friends on more than one occasion to give a sort of AI orientation because I’m afraid of seeing them get left behind. — Claire
The number of young adults who simply cannot be bothered to sit down and read a single page of text is terrifying. — Charles
And finally:
If you think things are weird now, you haven’t seen anything yet. — Tom
Overall this is a brilliant list that makes me feel less alone in the world. The part about "too much productivity" and burnout particularly hits. I'm starting to feel stretched thin and frantic, despite getting absurd amounts of things done each day. Just one of the many wise and useful things here.
BUT....I think I disagree with every single thing under "On winners and losers in the AI era". Cloudfare and Stripe are just good utilities. Analytics are nothing - insights are everything - and AI is bad at working with numbers (without a lot of help) and needs a human with context and experience to get insights. Apple is bloated and now in their shareholder-beholden innovation and value death spiral - they just don't know it yet. And fair use - it's too messy to sort out. If a human can read something, recall it, and use it - an AI can. You can't put tht genie back in the bottle.