dowhy
DoWhy adds one estimation method a year and keeps its identification edge.
A side-by-side editorial comparison of Hex and iris — release velocity, themes, recent moves, and the top alternatives to consider.
| Feature | Hex | iris |
|---|---|---|
| Sector | Analytics | Analytics |
| Velocity score | 6.3 | 2.5 |
| Sparks · 30d | 1 | 0 |
| Top themes | agents, cli, mcp, model-routing | earth science, release cadence, python, release candidates |
| Last editorial update | 1d ago | 1h ago |
| Website | — | Visit → |
Hex is moving its agent out of the notebook and onto programmable surfaces.
Hex ships a fortnightly digest, and the through-line across the last ten is agent reach. The agent now runs from a CLI and API, appears inside Codex, acts as an MCP client against the user's own apps, can search the web, and can see the generative apps it builds. Alongside that, the model layer has become a customer-facing control: a picker, an admin-set default, Auto with visible attribution, and a rotating roster including Fable 5, Kimi K2.7, Claude Opus 5 and GPT-5.6.
Iris ships steadily on a two-a-year cadence, but its feed publishes only pointers.
Iris tags a release candidate roughly every four to five months — 3.13 through 3.16 over the past year — and the cadence is the only thing the feed actually reports. Every entry is the same seven-line template: a line saying this is a release candidate, conda-forge and PyPI install commands, and a link to a 'What's New' page held elsewhere. No release notes reach the feed at all.
Hex ships a fortnightly digest, and the through-line across the last ten is agent reach. The agent now runs from a CLI and API, appears inside Codex, acts as an MCP client against the user's own apps, can search the web, and can see the generative apps it builds. Alongside that, the model layer has become a customer-facing control: a picker, an admin-set default, Auto with visible attribution, and a rotating roster including Fable 5, Kimi K2.7, Claude Opus 5 and GPT-5.6.
Hex is positioning the agent as a component other systems call rather than a feature users visit. Governance is being built out in parallel — spend limits, credit usage controls, enterprise role-request controls, signed embedding — which is the pattern of a product preparing for programmatic usage it does not directly supervise. The generative-app surface is drifting the same way, from generated output toward editable, brandable, embeddable artifacts.
Expect the CLI and API surface to deepen before the notebook UI gains much, with more of Context Studio and app management reachable programmatically, and expect the spend and role controls to keep pace as headless usage grows.
Iris tags a release candidate roughly every four to five months — 3.13 through 3.16 over the past year — and the cadence is the only thing the feed actually reports. Every entry is the same seven-line template: a line saying this is a release candidate, conda-forge and PyPI install commands, and a link to a 'What's New' page held elsewhere. No release notes reach the feed at all.
The version numbers say a mature Met Office library is being maintained on a predictable schedule; nothing in the published entries says what is being maintained. Until the project puts release content in the tag body, its public trail will read as cadence without substance, and readers have to leave the feed to learn anything. The pattern has been identical across four consecutive releases, so it is a deliberate publishing choice rather than an oversight.
Expect v3.17.0rc0 around late 2026 on the same schedule, carrying the same boilerplate — the notes will again live on the documentation site rather than in the release entry.
Other Analytics products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either Hex or iris.
DoWhy adds one estimation method a year and keeps its identification edge.
OpenHouse is hardening the seams where table policies and jobs quietly fail.
silx 3.0 moved its default Qt binding to PySide6 — a migration for everyone embedding it.
statsmodels ships only what the ecosystem breaks — six releases, no new statistics.
StatsBase.jl is in caretaker mode — correctness fixes in, dependency bumps out.
Shiny made reactive apps observable, then gave them a way to tear themselves down
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
They serve adjacent needs but don't currently overlap on shipped themes. Hex is currently shipping more aggressively (velocity 6.3 vs 2.5), with 1 editorial sparks in the last 30 days against 0. See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. Hex is currently shipping more aggressively (velocity 6.3 vs 2.5), with 1 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other Analytics products to evaluate alongside.
Top Hex alternatives in Analytics are ranked by recent ship velocity. Browse the "Hex alternatives" section above for the current picks, or visit /alternatives/hex for the full list with editorial commentary on each.
Top iris alternatives in Analytics are ranked by recent ship velocity. Browse the "iris alternatives" section above for the current picks, or visit /alternatives/scitools-iris for the full list with editorial commentary on each.