iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Time Travel Coding is a planning-first workflow: describe the program in a Markdown file, refine that description with an AI coding agent, and only then ask it to implement the software. Michael Murphy presents this as a way to catch changes while they are still edits to a plan rather than rework in code. The idea is plausible, but its token savings have not been measured.
What “Time Travel Coding” means
In his September 30, 2026 DEV Community article, Michael Murphy argues that an agent can spend effort implementing a version of a program that the requester later changes. His proposed alternative is to use Markdown as an inexpensive place to explore what the program should do and feel before code is written. The software is the eventual result; the document is where the idea gets revised first.
Murphy sums up the approach as “Iterate the plan, not the program.” It is not a special Markdown format or a promise that an agent will understand every requirement. It is a sequence for making decisions about the intended product before implementation begins.
How to use the workflow
-
Describe the idea in plain language
Start a Markdown file with who the program is for, what it does, and how it should feel to use. Include the core purpose and experience you have in mind; do not begin by trying to specify every technical detail.
-
Ask the agent to picture the finished program
Use Murphy’s sample question: “Can you see what this looks like when it’s finished?” Ask the agent to describe the program screen by screen. This makes an abstract idea easier to inspect: you can notice missing screens, unclear steps, or assumptions about what a user sees.
-
Collect useful changes in the Markdown file
Ask what is missing, confusing, or could be improved. Decide which suggestions fit the product, then write those decisions into the plan. The point is not to accept every suggestion; it is to make the intended result clearer before implementation.
-
Consider how the idea might grow
Murphy suggests imagining what the program could look like if it continued growing at its current pace for 30 years. Treat this as a prompt for surfacing constraints or design choices that may matter later—not as a forecast, a deadline, or an instruction to build every imagined feature now.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy. -
Revise until the suggestions lose value
Continue asking questions and updating the plan while the agent is surfacing meaningful gaps or improvements. Murphy’s proposed stopping point is when new suggestions become small or repetitive.
-
Implement from the refined plan
Once the planning pass has clarified the intended program, ask the agent to build it. The Markdown file gives the implementation request a shared reference for scope and experience; it does not replace reviewing the result.
Record visual constraints, then check the result
Murphy recommends writing down design rules that should not be broken. His examples include avoiding glowing gradients or nested cards, using one accent color, and including the real words on every screen. These are illustrative choices, not universal design rules. Replace them with constraints that suit your own product.
Rank #4
After implementation, his suggested check is to ask the agent to open the app in a browser, capture a screenshot, and compare it against the written rules. That gives you a concrete way to review visual direction alongside behavior. A screenshot check can reveal mismatches with the plan, but it is not evidence that the application is complete or correct.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWhat the method can—and cannot—show about token use
Murphy’s rationale is that plan changes are cheaper than rebuilding code and that a fuller plan may help an agent avoid wrong turns. His article does not report token counts, a cost comparison, sample size, measured productivity results, or a controlled experiment. There is no established percentage, dollar amount, or token figure for savings from this workflow. “Stop burning tokens” should therefore be read as an intended benefit, not a proven result.
Best Value
- Hardcover notebook with line-ruled pages (front and back); ideal for notes, lists, journaling, and more
- 240 pages
- Archival quality; acid free
- Expandable inner pocket for storing loose items
- Includes bookmark and elastic closure
Official vendor guidance provides only limited context. Anthropic’s Claude Code help guidance recommends considering Plan Mode or asking for a list of files and intended changes before implementation on work affecting multiple files. OpenAI’s Codex help guidance says usage depends on the model, execution setting, task complexity, context, reasoning, speed, and tools. Those points support planning before some coding work and recognizing that agent usage varies; they do not establish that a Markdown-first process reduces usage.
Quick Recap
When this approach is most useful
- Use it when the intended screens, user flow, or product feel are still unsettled and you want to explore those decisions before code exists.
- Keep the plan proportionate. The workflow is meant to clarify the intended program, not to turn a small request into an exhaustive future roadmap.
- When a task affects multiple files, a planning or read-only exploration step may also help make intended changes visible before implementation, as Anthropic’s guidance describes.
- Judge the process by whether it produces a clearer implementation target and catches mismatches earlier—not by an assumed token-saving rate.
Sources
- Michael Murphy, “Time Travel Coding: Stop Burning Tokens – Build It in the .md First,” DEV Community, September 30, 2026.
- Anthropic / Claude Help Center, “Models, usage, and limits in Claude Code.”
- OpenAI Help Center, “Using Codex with your ChatGPT plan.”
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
