Recommendation · MoneyRandy Hunt recommends that a specialist's first job should be building an agent that scales their own expertise and best practices — for example, a solo UX writer encoding their best practices so the whole company benefits while the specialist monitors and tunes it. He reports Notion's UX writing agent became people's favorite tool company-wide within roughly two months, and that colleagues also learned what the specialist does and when to reach out directly.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Recommendation · MoneyRandy Hunt recommends using AI research agents with access to historical research context to answer 'have we ever talked to customers about this?' in real time during the design process, and querying them for prior customer interviews and past findings on a topic before starting fresh. He discourages relying on the old workflow of asking the research team in Slack, waiting on a person who may have left, or an unfinished research repository.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Creator perspective · MoneyCarmen Li explains that to value a GPU asset, you discount the cash flows from a lease back to time zero — e.g., signing a three-year lease on A100s at a given rate gives you the asset's worth. The A100 residual value curve is downward-sloping (depreciating) but sits above the straight-line accounting depreciation curve. The H100 residual value is 'going bananas' — the price actually went up, analogous to secondary car prices going up during COVID. The B200 residual value has gone above MSRP by this point.
Valuation methodology and market observations from the speaker; the H100 analogy carries a possible numeric error flag.
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Report · MoneyCarmen Li describes Compute Exchange's focus on reserve compute, forward contracts, and token packages, covering three major layers of compute: reserve GPU, forward contracts, and open hardware market. Through the platform, you can get reserve contracts and forward contracts on both GPU and token. The minimum reserve contract is one node starting at one month, going up to three, four, or five years depending on availability and providers. If you overcommit in terms of quantity or duration, you can sell it back to Compute Exchange, providing flexibility.
Product descriptions from the founder; contract terms may vary by provider.
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Report · MoneyOn the Compute Exchange platform, users can browse marketplace listings, chat with providers, and figure out the best delivery mechanisms. They can select vendor type, GPU model, deployment type, number of GPUs, compute interconnect preferences, CPU platforms, storage, and optional items like systems, software stacks, and regions. If they don't see anything they like, they can submit an RFQ. Behind the scenes, Compute Exchange has inventory tracking with over 100 providers globally, knowing who has forward capacity. When a request is submitted (e.g., 300 B200 GPUs starting January 1st), it is routed to providers thought to have those resources, who respond through the platform with a rubric spec including pricing, specs, SLAs, and benchmarks. Price is not the only indicator; the platform's
Process descriptions from the founder; the routing example may contain a numeric error.
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Report · MoneyTenex pays engineers based on story points agreed in advance rather than time. Some 10x engineers make over a million dollars a year while others make less, purely based on productivity differences. There is roughly a 4x difference in earning capacity between the highest and lowest paid engineer at Tenex. CJ believes this is what the future of engineering looks like and that companies will increasingly pay based on outcome rather than time as AI solves tech problems.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · MoneyTenex measures whether its AI engineering processes are working by tracking the trend in average story points per month of every engineer in the company.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · MoneyThe other strategy in the market, which is what the labs do, is an FDE (forward-deployed engineer) approach: a human goes into your business, understands how it runs, then builds custom software with AI that uses the lab's AI. Enterprises are described as willing to spend around $5 million a year paying OpenAI for forward-deployed engineers.
the $5M figure is approximate ('like 5 million') and speaker-reported
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyRandy Hunt contrasts the bad version of a small team — too small, lacking bandwidth, pushed to the delivery end of a timeline, ending up in production-engine mode — with the better Notion version, where constrained bandwidth acts as a forcing function for prioritization and gives the talent-dense team outsized influence by participating from the beginning of making things.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Recommendation · MoneyTenex's recommendation to companies is to first build a process around the organization, then take steps to hand off that work to an agent in the background. The real leverage from agents comes when agents surface work to you, you review it, and you throw it back — becoming an agent manager.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Recommendation · MoneyThe team vibe is that Opus 5.5 is the preferred model: it is fast, cheap, and good enough to cover the full range of needs that previously required two to three different models. Run a two-model workflow using Opus 5.5 as the primary driver and Codex as a reviewer, working back and forth on your machine — this still produces the best results. Reach for Codex when the task involves computer use; keep Opus as the general default. When leaning on Codex, the model used is Sol 6.1. A lot of people on the team are defaulting away from Fable.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Recommendation · MoneyHook your Claude Code subscription into AMP to cut token costs by roughly 80% — AMP recently launched this capability and it has been a big game changer. AMP's cloud 'orbs' setup takes work off your local machine and can replicate your local workflow, including hopping to another orb to review another orb's output, which is why the speaker prefers it over cloud alternatives. The speaker is AMP maxing — using Opus through AMP.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Recommendation · MoneyRandy Hunt recommends intentionally evaluating whether an organization values design before joining it, rather than hoping you'll be lucky. He distinguishes funding design from knowing how to put it to work: an organization can budget for design without organizationally knowing how to value it, as he observed at Grab.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Recommendation · MoneyRandy Hunt recommends giving experienced team members both agency and responsibility and trusting them: ownership over shipped work drives satisfaction and retention, and people get a space to initiate, drive, push through, and get work out.
Design Better · Randy Hunt: Notion and Etsy design leader on why the best design teams stay small
Principle · MoneyCarmen Li argues that GPU behaves just like every commodity, with a very clear term structure risk that data futures are excellent at hedging. A lot of block trading has been happening around the GPU term structure.
The creator's argument, grounded in her trading background.
AI Engineer · Buying GPUs Should Work Like Buying Stocks — Carmen Li, Compute Exchange
Creator perspective · MoneyThe harder you move up-market, the harder it is to rip out an existing SaaS and rebuild from scratch — Salesforce and SAP are not realistically at risk. Before replacing a mid-tier SaaS tool in-house, ask whether the savings (as a percentage of revenue or cost) will ever move the needle enough to justify allocating engineer resources — the answer is case-by-case. Decide whether to build software in-house by asking: is the size of the prize big enough to justify the cognitive load of maintaining something not core to your business? Remember that part of software's value is the headaches you don't have to think about in maintaining and improving it — not just the utility you get from it.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Principle · MoneyAlex Lieberman explains Tenex's recommendation to companies: first build a standardized process around a function before handing that work off to an agent. When 40 people do the same function in slightly different ways, it is much harder to build an agent that proactively does that work in the background. Without standardization, it is impossible for an agent to know whether it is doing the work right, to build evals, or to deliver the value an agent performing that function would have. The real leverage comes when agents surface work to you, you review it, and throw it back — becoming an agent manager.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Report · MoneyCJ Hess describes the meta harness — the engineering workflows at Tenex that enable meaningful work on long-running tasks. It grew from an initial effort to standardize a set of skills into a structured set of documentation within a repo plus a CLI that models use to drive the work, including creating plans, visualizations, and client readouts. The harness builds all the pieces of the process around actually writing code so it happens more automatically and standard across the team.
Latent.Space · AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hess
Creator perspective · MoneyThe unit of value in software has changed from the app to the computer — meaning from an app to anything a human can do with a computer — abstracting away the app layer entirely, and value will drop in the app layer.
speaker's thesis, described with 0.8 confidence
The a16z Show · What Happens When Al Agents Can Spend Your Money?
Creator perspective · MoneyDylan argues the main difference between running a model and running an actual production AI system is the full loop from inference to training to deployment: running models in production is complicated and takes work, teams need to understand how a model performs in the real world and improve it over time, and redeploy updates without breaking things. The full cycle from running to observing to capturing and redeploying separates teams that ship great AI products from teams that ship demos.
AI Engineer · What Makes Open Models Fast in Production — Sujee Maniyam, Nebius
