Principle · MoneyTenex's ethos is to reward engineers for their output as opposed to rewarding them for their time. The thesis is that AI is really just a leveraged bet on a human's intellect and work ethic, so there will be more variance than ever before in earning capacity. When AI is solving tech problems, companies will pay based on outcome rather than the time required to build something. Tenex's model is for the person who wants to go long themselves, has a strong software engineering foundation, and constantly pushes to leverage the frontier.
the speakers' thesis
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyIf you are a large enough company, the labs won't train on your data, but if your way of operating your business makes it into the weights of a model resold to others, you are effectively selling what makes your business successful to someone else — and you really don't want to do that. The speaker's company does not train on customer data; it is not part of their philosophy as a business.
speaker's heuristic and statement about their own company's policy
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Principle · MoneyYou want a neutral agent that is yours, where the only source of money for the agent is your wallet as the user — so its incentive is to give you the best information with no distortion from marketing forces. You can still consume marketing intentionally: ask your neutral agent 'what are people talking about on Instagram?' or 'what's being advertised on it?' — there you want to be flavored by advertising, but the agent itself stays neutral vis-à-vis that during purchasing decisions.
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyAlex Lieberman and CJ Hess are generally bearish on vertical-specific AI solutions in different functions or industries. Spending all day in the guts of enterprises reveals enormous nuance, inertia, systems to wrangle, and people to wrangle. Vertical AI products where an agent is the product doing work in a specific vertical will struggle going forward unless there is a strategic compliance or data moat, as the ability to spin up your own agent quickly compounds and models get better and tasks fall into general model capability. Vertical products that change the process and have an agent on top of an app and tool built around it can still be super helpful and successful.
the speakers' own interpretive stance
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · MoneyOf the roughly 100 enterprises Tenex talks to, 90% have rolled out Claude or Codex to employees, and training/enablement is at roughly 60%. Most people guess 70-80% of enterprises have rolled out coding agents to the engineering org, but the actual number is closer to 50% — every other enterprise either hasn't rolled out a coding agent or what they use is closer to not having one. The biggest diffusion gap is agents in production for non-engineering work, which drops to like 5%; most enterprises with more than 1,000 employees do not have an agent in production doing work that isn't building software.
sample is enterprises Tenex speaks with, not all enterprises
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyData ownership and neutrality are so important that large companies should not be beholden to any one lab, especially given labs are building vertical companies that compete with their customers (e.g., Rogo, Harvey).
speaker's strategic argument, endorsed
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyThere aren't enough tinkerers in the world — or people with time to be tinkerers — to help their businesses, and that is why AI is not diffusing. Silicon Valley firms use AI effectively because everyone works next to engineers and tinkerers, while most of the world lacks that proximity. If you can make a non-tinkerer get the same value as a tinkerer in the world of AI, you could be a ginormous business today because AI can solve so many business problems.
speaker's theory and argument
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyIf you can make a non-tinkerer get the same value as a tinkerer in the world of AI, you could be a ginormous business today — the intelligence is already there; the problem is deploying it effectively at scale.
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Recommendation · MoneyCarmen Li recommends performing KYB/KYC verification on clients before extending compute credit or reserve contracts, to ensure they can pay at a scheduled rate.
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Method · MoneyTo build role-specific automation recommendations without violating user privacy: identify power users via aggregate metrics, interview them, extract the 5 things they have in common, then hand-build prototype 'router' automations as seed templates for new users in that role.
described
Method, not a verified protocol.
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Recommendation · MoneyRandy Hunt advises putting five or six incredibly good people around each other, especially if they have reasonable control of their ego, because really good people thrive around other really good people.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Recommendation · MoneyRandy Hunt recommends setting team expectations that there is no internal ladder — no titles, no levels, no training-and-career-path framing. People join to make things happen and are compensated fairly for it.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Recommendation · MoneyRandy Hunt recommends practicing 'whole design': aspiring to design every single touchpoint and angle of the experience, not just the core product surface. A push notification is just as much of a design opportunity as a core screen, and the full journey spans how people discover the product in social, functional and service emails, and push notifications that pull users back into the core experience.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Creator perspective · MoneyThe BlackBerry died quickly because every CEO got an iPhone for Christmas, experienced how good it was personally, and then told their teams to make it work at their company. Steve Ballmer said nobody would buy an $800 iPhone and that it would be a flop. Good Technologies was an enterprise-ready email client app for the iPhone that the speaker describes as the ugliest thing in the history of the world; the speaker used it. Being a Good Technologies-style enterprise overlay on a consumer device is not a good business because people judge you against the quality level of every other app on that device. The risk for the speaker's company is not just being the BlackBerry but also being Good Technologies — an enterprise overlay that loses to the native consumer app that combines convenience and
the BlackBerry theory is explicitly the speaker's own
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneySujee explains that LLMs generate tokens one by one and large models take a long time to generate these tokens. Speculative decoding employs a smaller model, which is a lot of the time faster and cheaper, to generate the tokens, then has the large model verify at the end — like a senior engineer farming out work to a junior engineer who does the heavy lifting, and then the senior verifies the work. It surprisingly scales really well. If the large model doesn't like the answer, it will regenerate the tokens — that's the worst case scenario. Most of the time you get pretty good performance by mixing in small models and large models.
Also discussed by Dan (speaker, name from Q&A) on Y Combinator, Y Combinator (unclassified source — not counted until classified). Not counted as independent sources.
AI Engineer · What Makes Open Models Fast in Production — Sujee Maniyam, Nebius
Recommendation · MoneyCarmen Li describes a risk scenario: companies with servers on their balance sheet and five-to-six-year financing obligations but reserve contracts covering only the first three years are exposed in years four through six, and should hedge that risk via swaps or block trading. She also notes that block trading is happening even within the first three years of reserve contracts, hedging against the risk that a reserve contract defaults.
partially hedged (described practice)
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Recommendation · MoneyAt least once a quarter, sit down and re-evaluate your story point criteria and how you're estimating projects. As models get better, you need to calibrate the value being created rather than only rewarding raw output — there's a fine balance between rewarding output and recalibrating its worth. In theory, story points should deflate as models improve (a single story point is not worth the same now as a year ago), though the team is uncertain about the actual direction of inflation/deflation.
partial
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Recommendation · MoneyWhen marketing AI capabilities, be concrete about the outcome, not the feature: 'TOWN now has computer use' means nothing; 'I can make posters for this event that look just like the ones you've been making in Canva, using your Canva subscription' gets a yes.
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Recommendation · MoneyDistinguish between friction you enjoy (browsing, exploring, shopping in person) and friction you don't (reordering routine items), and delegate only the friction you don't enjoy to an agent. Don't optimize every AI agent interaction for maximum efficiency — humans deliberately choose friction and entertainment in some contexts (browsing, Instagram, in-person shopping). When buying routine household items like toilet paper, use an agent to get the same item as quickly and cheaply as possible. Accept ads on platforms where you're browsing for discovery (like Amazon when shopping for kids' toys) — an unbranded interesting product can lead to asking your agent for reviews and a purchase you wouldn't have made otherwise.
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Recommendation · MoneyAn agent onboarding pattern that converts: do one valuable task for the user cheaply (e.g., a first pass for 25 cents), then proactively suggest recurring automation — 'do you want me to do that before every event like this that I noticed in your calendar?' — and as models improve, proactively suggest increasingly impossible-seeming capabilities, e.g., 'do you also want me to make video assets so you can advertise this on Instagram?', just following the curve of what models will be able to do.
the 25-cent figure is flagged as a possible numeric error
The a16z Show · What Happens When Al Agents Can Spend Your Money?
