Episode 1: Terralith: the Future of Terraform Automation and Collaboration? (Re-Upload)
Published: Saturday, Oct 4, 2025 • Duration: 47 minutes • Season 1
Download MP3 | Watch on YouTube
Vincent De Smet is working on https://terraconstructs.dev/
Opentaco trailer from https://www.linkedin.com/posts/izalutski_today-i-am-excited-to-announce-diggers-biggest-activity-7376635476264374272-O2Cp/ (without the copyrighted aggressive music)
summarize "https://youtu.be/UXIJe2moZ-I" --timestamps --slides
This video is a conversational deep-dive into recent turbulence in the Terraform ecosystem: licensing shifts, forks, new orchestration products, and a very public community spat over an “open standard” play. The hosts map how teams currently run Terraform (single repos, many states, or cloud stacks), explain why multi-state orchestration is hard, and trace multiple vendor responses: Atlantis (open source), runner-based alternatives like Digger, managed vendors (Spacelift, M0), the Terraform Cloud “stacks” feature, and new projects pitching state-centric tooling (stategraph.dev and Digger’s “OpenTaco”). The conversation includes concrete operational trade-offs: why teams split states (ownership, lifecycle, blast radius), the risk of race conditions between resources (API Gateway example), how Atlantis’ PR lock + approved-plan workflow works, and why many orgs are wary of resource-under-management pricing versus self-hosted tooling. The hosts quote numbers and anecdotes from real clients: small self‑hosted ops can run on a ~$15 EC2 instance, many teams pay ~$20/month for simple setups, and some cloud quotes for managed “resource under management” tooling can balloon to ~$10,000/month for very large estates. The conversation highlights two recurring themes: Context is king. and the backlash topic that motivated the heated Reddit/Hacker News threads: resource under management pricing is criminal. The summary below follows the video’s slide timeline and gives a compact headline plus one-line summary per slide so you can judge whether to watch the clip.
Ecosystem context
Hosts set the stage: Terraform’s commercial relicensing led to forks (OpenTofu) and many vendors building “tacos” (Terraform Automation/Collaboration Software) such as Spacelift, M0, Atlantis, and Digger; Digger’s pitch is using existing CI runners instead of running a dedicated server.
Orchestration problem and stacks GA
Terraform Cloud’s “stacks” (now GA) add cross‑state orchestration—helping sequence dependent states like network → cluster → GitOps controller—but this is cloud‑only and only partially surfaced to the CLI, so many orchestration gaps remain.
Dependency complexity and race conditions
Within a single state Terraform handles resource graphs well, but implicit dependencies across states create races (example: API Gateway deployment failing because an integration/lambda isn’t attached), and cross‑state outputs can change shapes, breaking dependents.
Atlantis vs runners vs pricing
Atlantis provides PR locking, plan-on-PR and apply-from-approved-plan workflows and can run cheaply on a small EC2 ($15) or ~$20/month setups; managed vendors charge per‑resource which can scale from trivial to staggering ($10,000/month quoted for very large fleets).
Monolith vs many states debate
Two camps: tools that advocate splitting state for lifecycle/ownership reasons (TerraGrunt/Gruntwork approach) versus proposals to treat a large single state (“teralith”) with better tooling (state graph / monolith tooling); hosts favor state‑graph ideas for richer tooling.
OpenTaco launch and backlash
Digger released “OpenTaco” (state manager, module registry, RBAC, dependency graph, insights) and a provocative video; the community pushed back on claims of an “open standard,” after the repo included copied Terraform Cloud backend code without attribution, triggering removal, heated Reddit/HN threads, and an apology post.
Model: openai/gpt-5-mini