Strategy Breakdowns replaced prompted graphics with a coded brand render engine

• Strategy Breakdowns replaced one-off image prompting with a custom, code-based generator that takes data and emits a finished branded graphic, as the build announcement describes.
• Exact fonts, the purple #B19EF8, bendy arrows and a “software-but-warm” feel are among the rules that the build announcement names.
• The workflow is accumulating a repository of 37 graphics, with more shipping daily according to the build announcement.
• The same publisher is also applying account-specific automation to its sponsor kit, where a Ploy post claims eight customized sites were made in about 20 minutes.

A short demo moves from data entered in a custom interface to a finished Strategy Breakdowns visual. A separate sponsored Wispr Flow post turns meeting context into a six-step diagram, while the Ploy post lays out a self-updating website loop.

Strategy Breakdowns says in the build announcement that it commissioned Hourglass to build an MVP render engine. Data goes in, then a finished branded graphic comes out.

That is a narrower job than “make an image.” The system was built around the publisher’s specification and refinements, after manually producing every graphic became too slow. A follow-up post directs readers to Hourglass, whose official site describes the company as an Australian AI implementation shop that builds tools and agents for clients.

The build announcement lists the render target:

• the right fonts
• the exact purple, #B19EF8
• bendy arrows
• a “software-but-warm” visual tone

The publisher says image-model attempts failed on charts, color matching and repeatability. Fixing those bad generations “was eating all our time,” it said in a reply, before summarizing the change as another reply did: stopping prompts and starting a build.

The engine’s value is repeatable constraint satisfaction. Its visual language lives in code rather than surviving one lucky prompt.

The team says it already has 37 graphics in the repository and is adding more every day in the build announcement. The post calls them “graphics,” not templates, so it does not say how many distinct layouts power the set.

That distinction is useful: the proof here is a production archive of rendered outputs, not a single polished demo.

A paid Wispr Flow post from the same publisher is not labelled as an output of the new render engine, but it shows the kind of information-dense branded explainer the team publishes.

Its Notetaker sequence is:

1. Brief before the call
2. No bot joins
3. Live transcript and catch-ups
4. Real names, terms and speakers
5. Verified summary after
6. Feed the result to AI through MCP

The post says the service captured its newsletter jargon and acronyms, then passed the transcript into Claude through MCP for routine task assignment, deal progression and behind-the-scenes content.

A different Ploy post describes the publisher applying the same appetite for variant production to a sponsor website, rather than to the graphic engine. It says the kit has startup and enterprise versions with different copy and logos, identifies and scores visitors into a CRM, and automatically tests copy and calls to action.

The publisher also says it made eight sites tailored to target accounts in about 20 minutes. A follow-up reply adds the reason for splitting the messaging: sponsor kits need different angles for companies of different sizes.