fun while it lasted... predictions on how long til Fable returns?

Enterprise customers pull you in different directions. The hard part isn't picking which nudges to follow, it's finding the product shape that none of them asked for, but all of them need.

building software is the thing i genuinely can't stop thinking about. not the business side, not the fundraising, not even the wins. just the craft of getting to "holy shit, it works" on that didn't exist before. that moment never gets old.

Quite a bit more coming from @tembo team in June :) Excited to share.

if you engage a little w nba content on x watch out for timeline takeover

enterprise is mandating "4x developer output" but can't even measure current output anyone figure this out?

Your agents start broken. The question is whether your system is designed for them to mature or whether you just pretend they ship ready on day one.

trust me - linear @tembo is best effort to results ratio

darren and the tembo eng team rn iykyk

Every SaaS company bolting a coding agent sidecar onto their core product is building a 2025 demo they won't be able to maintain in 2027. A coding platform is a "startup-sized" problem.

tembo team’s in NYC next week who wants to grab a coffee / who should we meet?

we intentionally never supported bringing your own claude code sub in @tembo from the beginning still possible to bring your own codex subscription though and then only pay for the VM compute costs

Enterprise customers rarely churn because your product is “bad.” They churn because you left the exits wide open: shallow integrations, generic workflows, easy data portability, and no real dependency on outcomes. The moat isn’t flexibility, it’s becoming the operational layer they can’t rip out: embedded in processes, wired into systems of record, trusted for compliance, and relied on for daily decisions. Make leaving costly in time, risk, and lost leverage, not in contracts.

Coding agents were step one. Now we're building agents for GTM, ops, finance, support, and every other knowledge-work function—agents that research, draft, route, reconcile, and ship decisions at speed. Each one captures the playbooks, closes the loop with humans, and gets smarter from outcomes.

Every operations leader I talk to has the same problem. They bought the CRM. They bought the ticket system. They bought the data tool. And their team still runs on post-it notes and Excel. The instinct is always the same: get disciplined, pick one system, force adoption, build dashboards. It never works. Not because the tools are bad, but because knowledge workers don't live inside any single tool. They live in Slack. They live in their inbox. They live wherever the conversation is happening. Instead of dragging people into the system of record, build an agent that goes to every system of record on their behalf. It pulls open tickets from your support platform. It pulls account ownership and last-touch dates from your CRM. It pulls revenue data from your analytics layer. Then it synthesizes all of that into a prioritized daily view, delivered where people already work. The key is that the prioritization logic isn't a black box. It's inspectable code. Anyone on the team can ask the agent to explain how it ranked their accounts. Leadership can review the algorithm and adjust the rules. When someone on the team has an idea for how priorities should change, that becomes a concrete improvement to the agent's logic, not a suggestion lost in a meeting. This is what enterprise AI operationalization actually looks like. It's not replacing your systems. It's not another dashboard. It's an agent that reads five tools at 7am so your team knows exactly where to focus at 8am. The data stays where it lives. The work gets done where people already are. The organizations that figure this out won't have better tools. They'll have the same tools, with an agent layer that finally makes them useful.

The gap between "automation" and "agent" isn't branding. It's architecture. Automations run on triggers. Agents run on context. Name things wrong and you build the wrong primitives.

Anthropic is hunting enterprise workflows to productize. The contrarian bet: a probabilistic system will never match an agent built by someone who deeply understands how their specific company works.