Start a product as horizontal as possible, let the market reveal which verticals matter, and only verticalize after learning from real users — don't presume you already know the niche problem.
endorsed
Distinguish the customer (who writes the check) from the user (who deals with the product); building only for the customer risks making the CISO happy while the user suffers.
endorsed
Build products for the end user, not for corporate buyers (lawyers, 'corporate overlords'), if you want a product designed for the user — then work backwards into enterprise/sales-led growth later.
endorsed
Bet on AI improving to customize the product per user as the differentiation strategy when raw capabilities converge — recommendations tuned to the user's specific role (electrician vs VC vs EA).
endorsed
When product-market fit is strong and growth is organic, focus on making the product better before building enterprise admin features; defer compliance shaping until later.
endorsed
Target SMB and mid-market with self-serve product growth while building enterprise security/compliance/finance features for large companies.
endorsed
Estimate TAM by asking what infrastructure most businesses already run on — email and calendar are universal, especially for external-facing roles (sales, business development, recruiting), and it's global, not just US.
described
Estimate TAM by looking at how businesses communicate externally: if your product touches email and calendar, the market spans sales, business development, recruiting, and every geography.
described
Start with a world-class consumer product experience and adapt it for enterprise, not the reverse — it's almost impossible to start from compliance/IT requirements and then make the product experience incredible.
endorsed
Avoid businesses with infrequent purchase cycles — if customers only buy once a year, you can't build a DAU tool, word of mouth fades before the next purchase, and hyperscaling becomes nearly impossible.
discouraged
Define your market broadly as helping people find opportunities to deploy AI at their job, rather than narrowly as 'assistant' which casts too narrow a brush.
endorsed
When deciding whether to continue or pivot a startup, evaluate against your core motivations: (1) impact — can you still affect millions of people, and (2) team — can you look your employees in the eye given their opportunity cost?
endorsed
Randy Hunt recommends that design teams build momentum and relationships ground-up rather than trying to deliver design value from the top down.
Randy Hunt recommends getting influential collaborators to advocate for design on your behalf rather than pitching design's value directly, building ground-up working partnerships instead of trying to change design's perception through top-down communication.
Randy Hunt recommends reducing friction between making something and getting it into the world — the experience of shipping and having people like it is very sticky for retention — and when something doesn't land, waking up the next day and making it better.
Sujee recommends that for customers, open models mean choice: you can run open models in any provider with no vendor lock-in and save a lot of money on tokenomics.
Carmen Li describes Compute Exchange as a neutral marketplace for GPU procurement, launched in 2024, which she calls the world's only such neutral marketplace. The platform is vendor-agnostic and does not do on-demand compute.
The 'world's only' claim is the speaker's own description and may contain a numeric error.
CJ Hess says that at least once a quarter Tenex sits down and evaluates its criteria for story points and how it is estimating projects. He imagines that in the future Tenex may look more at true output — how it works and how it looks — as opposed to the traditional story point model, while still anchoring to story points as the unit. He adds that there is a fine balance between still rewarding output and calibrating the value being created as models get better.
future shift is speculative ('I imagine')
The second common enterprise need is permissioning and RBAC — as every employee has Claude or Codex and needs access to certain data. Giving every employee access with certain data needs turns data permissioning into an absolute nightmare for CIOs and CISOs, and enterprises lack a good gateway for easily doing a control panel for data permissioning (what you can and cannot read).
There are limits to the story-point pay model that Tenex knew would be limits. Classic concerns about output-based pay are that story points are subjective, people will sandbag story points, or ship bad code that barely checks the box. Those two concerns have not been concerns at all because of Tenex's AI strategists, who are incentivized perfectly at odds with engineers — strategists are incentivized for customer retention and growth. The biggest challenge has been compensating engineers for all the time not spent coding, like architecting and meeting with clients; Tenex had to find nuance in the comp model for non-coding time, but it's still largely driven by output.
the host's self-assessment
No published protocol matches this selection.
The creators behind most of what cleared.
195 cleared insights match
99 cleared insights match
85 cleared insights match
Choose sources → to get a library of your own and a daily brief.
