MyShell VS Transpose

Transpose's answer:

Two things, and the second one is the harder problem.

First, direction. Almost every repurposing tool moves content toward LinkedIn, because that's where B2B budgets are. Transpose goes the other way. It takes LinkedIn posts and Reddit threads and turns them into native X content: single tweets, threads, and quote cards. Reddit as an input source is close to untouched by other tools, and it's the richest one, because real Reddit threads contain the specific detail and lived experience that makes a tweet land.

Second, voice. Transpose is built around the assumption that anything a reader clocks as AI-written is dead on arrival. So the constraints are baked into the product rather than left to the user: no em-dashes, no "in today's fast-paced world," no run of five identically-shaped sentences, specific details instead of vague claims. The output is meant to read like someone who works in your niche wrote it, not someone describing that niche from the outside.

Most tools in this space treat "sounds human" as a bullet point. For Transpose it's the whole product.

Transpose's answer:

Database and storage: Supabase (Postgres). Authentication: Google OAuth. AI: the Anthropic API, with Claude handling generation. Content ingestion: Apify for pulling source posts from LinkedIn and Reddit. Payments and billing: Paddle, acting as merchant of record. Transactional email: Resend. Analytics: Vercel Analytics.

The stack was chosen so one person could run all of it. Paddle in particular does a lot of quiet work here, since it handles global tax and merchant-of-record duties that would otherwise be impossible to manage solo from Ghana.

Transpose's answer:

Because most of the alternatives are solving a different problem.

Schedulers like Hypefury and Buffer solve distribution. They assume you already have the writing. Ghostwriter tools like Postwise and Tweet Hunter generate from a topic or a rough idea, which means you're still supplying the thinking, and their output has a recognizable house style that regular readers start to notice.

Transpose starts from something you already wrote or something a real person already said. The insight, the specifics, the actual substance are already there. The job is translation, not invention, and translation is a job AI is genuinely good at.

Then there's the free option: pasting a post into ChatGPT with a prompt. That works, sort of. It also produces the exact cadence and vocabulary that people have been trained to scroll past over the last three years. Transpose spends its effort on the part a generic prompt doesn't handle, which is stripping the tells.

Practical reasons too. There's a free plan with no card required, paid starts at $9 a month, and you can test the output quality in under a minute before deciding anything.

Transpose's answer:

Four groups, roughly in order of how much they need it.

Ghostwriters and content agencies who manage several client accounts and are already producing LinkedIn content that could be earning a second audience on X. They feel the time cost most acutely and they're the most sensitive to output that sounds artificial, because their client relationship depends on it.

Founders and operators growing a personal brand on X. Usually posting on LinkedIn already, usually sitting on a year or two of posts that never got a second life, usually short on time.

Content marketers running multi-platform distribution, who know a straight cross-post reads like it wandered in from another website but don't have hours to rewrite everything by hand.

Creators and newsletter writers who mine Reddit for real stories and want to turn them into X content without losing the texture that made them worth reading.

Common thread: they all already have the ideas. The bottleneck is turning them into something native to X that doesn't read like a machine wrote it.

Transpose's answer:

I kept noticing the same failure. Someone writes a LinkedIn post that does genuinely well, pastes it into X, and it lands on nothing. Not because the idea was bad, but because the writing was shaped for a completely different platform. LinkedIn rewards long, formatted, slightly formal. X rewards the opposite. A copy-paste is legible as a copy-paste within about two seconds.

I started building it with no development background. Every part of it was learned while shipping it, from the first line of code through to picking a payments provider, which turned out to be its own education. Stripe doesn't work for a founder in Ghana. Lemon Squeezy rejected the store. Paddle was the third attempt and the one that worked. That kind of thing eats weeks and nobody writes about it.

The direction that kept correcting itself during the build was voice. Every time the output got more capable, it also got more obviously artificial, and obviously artificial content is worthless on X no matter how good the underlying idea is. So the constraint became the feature. Transpose is essentially a repurposing engine wrapped around a very strict set of rules about what AI writing is not allowed to sound like.

Built solo from Accra, Ghana. Still solo.