2026-04-03 - DataHub One-Post Prototype

2026-04-03 - DataHub One-Post Prototype

Historical prototype. The Discord fetch script cited below was retired from this repo on 2026-09-27.

Summary

This was a narrow Volume and Velocity prototype run focused on shipping one real post end to end from an internal Discord update.

The goal was not to design the whole system. The goal was to prove that the existing setup can already support a fast human-in-the-loop workflow for one post:

  • fetch a real update from Discord
  • choose one promising candidate
  • turn it into a short working brief
  • use the existing social drafting skill
  • get ready-to-publish output

Result:

  • the process worked smoothly
  • no major friction was identified in this run
  • one of the generated LinkedIn options could be published immediately with little or no editing

Why This Experiment

This prototype followed the principle:

  • one working end-to-end post is more valuable than a broad plan that is not yet implemented

It used the existing tooling instead of creating new tooling:

  • volume-and-velocity/discord-inbox/fetch.py
  • volume-and-velocity/skills/fast-social-orchestrator/SKILL.md

Steps We Took

  1. We narrowed the goal to one real post instead of the full automation system.
  2. We used the existing Discord fetch script against the currently working channel.
  3. We reviewed recent fetched messages and selected one strong candidate.
  4. We chose a DataHub product update from 2026-03-20 about Observable Plot support and individual view pages.
  5. We manually converted that raw update into a short structured brief.
  6. We used the fast social orchestrator approach with tighter constraints:
    • LinkedIn only
    • 3 options
    • informative, concrete, practical, engaging
    • no strategy writeup
    • no extra questions unless blocked
  7. We reviewed the output and confirmed that at least one option was immediately publishable.

Candidate Chosen

Chosen update:

  • Product: DataHub
  • Date: 2026-03-20
  • Source: Discord #whats-cooking

Reason for choosing it:

  • clear product
  • clear shipped improvement
  • concrete URLs
  • easy practical value
  • strong fit for DataHub positioning around better data publishing and usability

Exact Brief Used

# Prototype Brief: DataHub LinkedIn Post

## Goal

Create one strong, ready-to-publish LinkedIn post for DataHub / Datopian based on a real internal update.

The post should be:
- informative
- concrete
- engaging
- practical, not hypey
- close to ready to publish with minimal edits

## Product

DataHub

## Raw Source Update

Source channel: Discord `#whats-cooking`
Date: 2026-03-20
Author: anuveyatsu

Raw messages:

1. We now support Observable Plot for visualizations:
https://datahub.io/ai/epoch-data-on-ai-models

2. We also have individual view pages now which makes it easy to embed views:
https://datahub.io/ai/epoch-data-on-ai-models/v/0

3. Added a new visualization here:
https://datahub.io/core/eu-emissions-trading-system

Context note:
A follow-up message saying “Ignore it please, there is no data for 2024” is not part of the public post angle and should be ignored.

## What Happened

DataHub now supports Observable Plot visualizations and individual view pages, making it easier to create, share, and embed dataset views.

## Likely Why It Matters

This is not just a cosmetic feature update.
It improves how published data can be explored, shared, and embedded, which reinforces the broader DataHub story around better data publishing and better usability.

## Desired Angle

Focus on practical value:
- better data publishing
- easier sharing and embedding of views
- more useful dataset exploration
- publishing data as something usable and explorable, not just a file dump

## Tone

Use the DataHub / Datopian tone:
- clear
- practical
- specific
- data-geeky in a good way
- no hype
- no inflated claims
- make the value legible quickly

## Channel

LinkedIn only for this prototype.

## Output Request

Write 3 LinkedIn post options:
1. safest
2. strongest
3. most engaging

Constraints:
- concise, but with enough context to be useful
- no hashtags unless truly necessary
- no thread or multi-post structure
- no strategy explanation
- no extra questions unless blocked
- each option should feel ready to publish

## Context URLs And Sources To Use For Enrichment

Use these first before making assumptions:

1. Product / stream context in repo:
- `volume-and-velocity/context/product-streams.md`
- `volume-and-velocity/context/voice-and-tone.md`
- `volume-and-velocity/context/channel-guidance.md`

2. Live URLs from the update:
- https://datahub.io/ai/epoch-data-on-ai-models
- https://datahub.io/ai/epoch-data-on-ai-models/v/0
- https://datahub.io/core/eu-emissions-trading-system

3. Related product context if needed:
- `offers/datahub.md`
- `north-star.md`
- `dashboard.md`

## Important Constraints

- Do not invent technical details not supported by the source or linked pages.
- Prefer one strong practical angle over trying to mention everything.
- Treat this as a real shipped improvement, not as a vague product announcement.

Outputs Generated

Three LinkedIn options were generated:

  • safest
  • strongest
  • most engaging

Safest

We’ve been improving how datasets get published and explored on DataHub.

DataHub now supports Observable Plot visualizations, individual view pages for easier embedding, and new dataset views like the EU ETS example.

The practical goal is simple: make published data more usable. Not just downloadable, but easier to explore, share, and embed in real workflows.

Examples:
https://datahub.io/ai/epoch-data-on-ai-models
https://datahub.io/ai/epoch-data-on-ai-models/v/0
https://datahub.io/core/eu-emissions-trading-system

Strongest

Selected as the best option for this prototype.

Publishing data well should mean more than uploading a CSV.

On DataHub, we’ve added Observable Plot visualizations and individual view pages, so dataset views are easier to explore, share, and embed.

You can see it in practice here:
https://datahub.io/ai/epoch-data-on-ai-models
https://datahub.io/ai/epoch-data-on-ai-models/v/0
https://datahub.io/core/eu-emissions-trading-system

Small product changes like these matter because better data publishing is really about usability, not just access.

Most Engaging

A dataset is much more useful when people can actually explore it, not just download it.

That’s why we’ve been improving DataHub with Observable Plot visualizations and individual view pages that are easier to embed and share.

A few live examples:
https://datahub.io/ai/epoch-data-on-ai-models
https://datahub.io/ai/epoch-data-on-ai-models/v/0
https://datahub.io/core/eu-emissions-trading-system

The direction here is straightforward: make published data easier to understand and easier to use.

Friction Observed

No major friction was observed in this prototype run.

What worked well:

  • the existing fetch script was enough to retrieve usable updates
  • the candidate was easy to identify
  • the manual brief was straightforward to write
  • the skill produced usable LinkedIn options quickly
  • one option felt ready to publish immediately

What This Suggests

This experiment suggests that the current setup is already capable of supporting a useful narrow workflow:

  • internal update
  • brief
  • skill-generated drafts
  • human review
  • publish

The next improvement should probably not be a broad redesign.

The next improvement should be to reduce the remaining manual glue around:

  • selecting candidate updates
  • assembling the brief automatically
  • repeating the process consistently on more real examples

Immediate Next Step

Publish one of the generated LinkedIn options and record:

  • whether any edits were needed
  • how long the process took in total
  • whether the source update had enough context on its own

That will turn this from a successful dry run into a real shipped prototype.

Edit this page
Built with LogoFlowershow