Migrating from d6tflow to oryxflow
The rename is the migration: your task classes, your API, and your cache all come with you.
Blog¶
Notes on making data analysis trustworthy and reproducible — for humans and AI coding agents. Results you can believe, lineage that says which code and inputs made every number, honest tool comparisons — and, because nothing recomputes twice, a trustworthy path that also happens to be the fast, cheap one.
Start here¶
- Why oryxflow — the positioning in one page: trustworthy results, reproducibility, lineage, and trustworthy AI data analysis, plus when not to reach for it.
- Reproducible data science workflows in Python — what makes a workflow reproducible, and the missing middle between notebooks and orchestrators.
Trust, reproducibility & caching
- Turn a messy notebook into a reproducible pipeline — migrate an analysis into tasks, one step at a time, so every result is tied to the code that made it.
- 4 reasons your machine learning code is bad — the failure modes a task DAG fixes, starting with the silent one.
- Stop rerunning your whole pipeline — how a change reruns exactly what it affects, so nothing sits on stale data and nothing recomputes twice.
- Cache intermediate DataFrames in Python — why hand-rolled pickle caches go stale, and caching by task identity you can actually trust.
Tool comparisons
- oryxflow vs the field — the full framework landscape for iterative, AI-assisted analysis.
- oryxflow vs Airflow — research workflows vs production orchestration.
- oryxflow vs Prefect — zero-config code-aware caching vs configurable orchestration.
- oryxflow vs Dagster — a lightweight research loop vs an asset platform.
- oryxflow vs DVC — native Python task identity vs file-hash data versioning.
- MLflow, or a reproducible pipeline? — experiment tracking vs workflow caching (and DVC).
- MLflow alternatives — the experiment-tracker landscape, and where a caching DAG fits beside (not against) it.
- Airflow alternatives for data science — research-loop tools vs production orchestrators, and when you don't need an orchestrator at all.
Trustworthy AI-assisted data science
- How to make Claude Code a trustworthy data scientist — where agents fail at pipeline work, and what removes those failures.
- Claude Code skills for data science — what a skill is, how it differs from a plugin, a slash command and an MCP server, and why it's the piece that keeps AI-written analysis reproducible.
- Best practices for AI-assisted data analysis — how to keep AI-generated analysis reproducible, auditable, and honest about correctness.
- The best Claude Code plugins and tools for data science — a by-the-job roundup of the AI data-science tooling landscape.
- The best AI tools for data analysis — the layers of the AI data stack, and the trust layer most of them miss.
LLM evals
- Your eval platform deletes your results in 14 days — eval platforms are working sets, not archives; keep the scored rows yourself.
- LLM evals are a parameter sweep — the scoring is new, the matrix underneath it is a solved problem.
- Cheap, durable LLM evals: pydantic-evals + oryxflow — how a metered eval matrix gets expensive, and what to cache instead.
Practical patterns
- Parameter sweeps without rerunning upstream steps — compare many configurations while shared steps compute once.
- Migrate from d6tflow to oryxflow — a whole-word package rename, not an API port; your cache comes with you.
- When not to use oryxflow — an honest guide to non-fit.
All posts¶
Cheap, durable LLM evals: pydantic-evals + oryxflow
An eval matrix is a Cartesian product. Metered platforms bill every cell, every time — then delete the answers.
Your eval platform deletes your results in 14 days
LLM evals are a parameter sweep. Use a parameter sweep tool.
LLM evals are a parameter sweep — use a parameter sweep tool
The scoring is genuinely new. The matrix underneath it is a solved problem from 2015.
Best practices for AI-assisted data analysis
AI coding agents write plausible analysis fast. The hard part — is it reproducible, and is it right? — hasn't changed. These practices are about making AI-generated analysis you can actually trust.
Airflow alternatives for data science: an honest roundup
Most people who search "Airflow alternatives" want a lighter orchestrator. Some of them don't need an orchestrator at all — they need a research loop whose results they can trust and reproduce. This roundup covers both.
The best AI tools for data analysis (and the trust layer most of them miss)
AI can write a whole analysis in seconds. The unsolved question is whether you can trust and reproduce what it just produced.
The best Claude Code plugins and tools for data science
The landscape is early, so stop shopping for a "best plugin" and start choosing by the job you need done — the one below keeps AI-generated analysis reproducible.
How to cache intermediate DataFrames in Python (without brittle pickle files)
Why features_v3_final.pkl keeps burning you — and how to make cached reuse something you can actually trust.
Claude Code skills for data science: what they are, and what they aren't
A skill is the part of a Claude Code plugin that teaches the agent how to work. For data science, that turns out to be exactly what's missing from "AI writes the analysis fast."