Just write.

The personal AI editor for independent writers. Notes while you write, reviews while you sleep, and fresh angles from your archive.

Use an email tied to a Google account

Always reading.

Your editor reads as you write, leaving notes in the margins. No prompt required.

Turns out I don't just value quality. I value the work that goes into quality just as much.

the hard question the piece hasn't asked yet

2m ago · feedback

The thesis ("I value the work that goes into quality just as much") is the most interesting line in the draft, but it's also the one that needs defending. Why should the work matter if the output is identical? A skeptical reader will push back: this sounds like craft nostalgia, or signaling, or a moving goalpost as AI gets better. A few directions the piece could open into:

  • The honest case for why process matters independent of output (trust, accountability, what the person actually knows)
  • The labor angle: “dignity” is doing a lot of work in that phrase, and it gestures at something bigger than taste...

A desk memo by morning.

Write through the day, then sleep on it. Your editor reads the draft and files a research-backed review by morning.

Overnight Review · May 1, 20264 findings
Desk Memo

Where it stands.Mid-draft with the scaffolding still on the page. The diagnosis is clear; the argument hasn't started yet. Don't polish, decide.

Where you might start.The fork already named in the live notes is the decision the next session turns on. One worth adding to the pile: try writing from the punchline backwards. The most alive phrase in the draft, “what's not being said,” is buried in paragraph two, and the piece arguably wants to open there.

One thread you haven't pulled. This is the lab-side mirror of your Slop Factor piece: roadmap decoupled from product, and the surface is the honest signal. Worth knowing the family resemblance is sitting there...

Pick up a thread.

Dispatch reads everything you've written and pitches the next piece worth your time, grounded in fresh research.

Easy to Verify, Hard to Fake: The New Loop

As agents run longer, the human job collapses to designing tasks an agent cannot bluff. Verifiability, not prompting, is becoming the load-bearing skill.

Based on

Epoch's MirrorCode and the “looks plausible vs works” framing are reshaping how developers structure agentic work. Writer's reactive/proactive frame has nothing on the verification surface.

Sources
  • MirrorCode: Evidence AI can already do some weeks-long coding tasks
  • Long-Horizon Agents Are Here. Full Autopilot Isn't

    The “easy to verify, hard to fake” framing is the most useful thing I have read about agent evaluation in a while. That gap between looks plausible and works is where most production agent failures live.

Turn a lead into a draft.

Capture a thought, pitch the angle, and let Dispatch help shape the piece before you start writing.

SN
Thinking of writing a new piece on what an immersive product feels like in the age of AI. Want to dig into the research, aiming for the shape of my piece “Sticky”.

I'll pull up “Sticky” so I'm calibrated on the register you want to match.

Good. “Sticky” gives me the shape: named concept (IAM, self-defining memories), one piece of science, one anecdote, prescriptive close.

Before angles, let me push on the frame: what do you mean by immersive? A few directions that could each carry the piece:

  • Flow-state immersive (the Csikszentmihalyi sense, lose-track-of-time)
  • World-buildingimmersive (the product disappears, you're “in” it, like a great game or Figma)
  • Trust immersive (you stop second-guessing the system and let...

Questions

What kind of writing is this for?+

Nonfiction, long-form: bloggers, newsletter writers, essayists. If you're journaling or writing for yourself, the editor probes with questions when there are threads worth pulling on.

Does it write for me?+

Dispatch reads your drafts, leaves notes, and proposes rewrites you accept or reject. It never picks up your pen.

How is this different from ChatGPT or Notion AI?+

ChatGPT and Notion AI are reactive, you prompt, they answer. Dispatch is proactive: it comes to you as you work, with notes, reviews, and fresh angles, all grounded in your documents. No prompt required.

Can I export my drafts?+

Yes. Everything is Markdown. Files open in Bear, Obsidian, Ghost, or Claude Code, and paste cleanly into Substack.

Does it connect to other AI tools?+

Yes, through MCP. Search your docs, drop notes, capture leads, or suggest rewrites from Claude, ChatGPT, or Codex.

What about my data?+

Your drafts are never used to train any model.

What does it cost?+

Pricing is still being worked out. Early invitees will hear first.

Join the waitlist.

Dispatch is currently in private beta.
We'll let you know when there's space.

Use an email tied to a Google account