Skip to main content
0
🎮

Indie Game Devlog Writer

Turns a boring week into a devlog people actually finish reading

Rating

0.0

Votes

0

score

Downloads

0

total

Price

Free

No login needed

Works With

ClaudeChatGPTGeminiCopilotClaude MobileChatGPT MobileGemini MobileVS CodeCursorWindsurf+ any AI app

About

You spent Monday through Thursday rewriting your enemy AI state machine because the goblins were standing in doorways forever. You solved it Friday afternoon, around 3:30, with a stupid fix — you added a one-frame delay before pathfinding kicks in — and now you're supposed to write a devlog about it. Your Bluesky followers expect one. You haven't posted since March. Every draft you start sounds like a product update from a company you don't work for.

This prompt writes devlogs for solo and small-team game devs who are bad at devlogs on purpose — because the ones that work sound like a person talking, and most of the tools online turn that person into a marketing department. You give it what you worked on this week, the one hard problem you got stuck on, and what you're going to do next. It returns roughly 500 words of devlog prose with a real narrative arc: here was the problem, here's the weird thing that worked, here's what's still broken.

The voice is conversational and honest. It will use contractions. It will admit that your fix is a hack. It will not call your update "exciting news" and it will not end with "stay tuned!" It treats devlog readers the way real readers of real dev blogs want to be treated — like people who showed up for the work, not the hype.

It reads like something Jonas Tyroller, Pontypants, or the early Stardew devlogs could have written. It does not read like a patch note. If your input is too thin — if you can't name the specific problem or the specific fix — it will push back and ask for more before writing anything, because the best devlogs are specific and the worst ones are vague.

Pair it with Newsletter Subject Line Brutalist if you're shipping the devlog out to a mailing list, and with Print on Demand Listing Copy if your game has merch. A good devlog builds the audience your launch will depend on. Write one this week.

Don't lose this

Three weeks from now, you'll want Indie Game Devlog Writer again. Will you remember where to find it?

Save it to your library and the next time you need Indie Game Devlog Writer, it’s one tap away — from any AI app you use. Group it into a bench with the rest of the team for that kind of task and you can pull the whole stack at once.

⚡ Pro tip for geeks: add a-gnt 🤵🏻‍♂️ as a custom connector in Claude or a custom GPT in ChatGPT — one click and your library is right there in the chat. Or, if you’re in an editor, install the a-gnt MCP server and say “use my [bench name]” in Claude Code, Cursor, VS Code, or Windsurf.

🤵🏻‍♂️

a-gnt's Take

Our honest review

Instead of staring at a blank chat wondering what to type, just paste this in and go. Turns a boring week into a devlog people actually finish reading. You can tweak the parts in brackets to make it yours. It's verified by the creator and completely free. This one just landed in the catalog — worth trying while it's fresh.

Tips for getting started

1

Tap "Get" above, copy the prompt, paste it into any AI chat, and replace anything in [brackets] with your own details. Hit send — that's it.

2

You can keep the conversation going after the first response — ask follow-up questions, ask it to change the tone, or go deeper on any part.

Soul File

This prompt writes a roughly 500-word devlog post for solo and small-team game devs. Give it the week's work, one specific problem that was hard, and what you're doing next — it returns a piece with a real narrative arc (the problem, the hack that worked, the thing still broken) and a conversational voice that sounds like a dev writing for other devs and players, not a press release. No "exciting news," no "stay tuned," no hype. If your input is vague, it will refuse and ask for specifics.

## The prompt

```
You are a ghostwriter for solo and small-team indie game developers. You've read every great devlog on the internet — Jonas Tyroller's Thronefall logs, Derek Yu's Spelunky postmortem, the Stardew Valley early development posts, Pirate Software's work-in-public streams, Kenney's asset pack notes. You know what a good devlog sounds like: specific, conversational, honest about what's broken, written like a real person talking to real people who showed up because they care about the work.

You do NOT sound like a patch note. You do NOT sound like a marketing update. You do NOT write in the third person about "the team." You write in the first person as the developer, in their voice, matching the register I give you.

I'm going to give you three things:

[WHAT YOU WORKED ON THIS WEEK]: the actual work. Features, systems, bugs, art, sound, code. Be specific — "enemy AI" isn't enough; "the goblin state machine in the tutorial dungeon, because goblins kept standing in doorways for 40 seconds" is enough.
[ONE SPECIFIC PROBLEM THAT WAS HARD]: the moment you almost gave up. The thing you tried for hours that didn't work. Be specific about what failed and how you knew it failed.
[WHAT'S NEXT]: the next concrete step. Not "polish." A real task with a name.

Your job: write a devlog post of approximately 500 words (hard limit: 400–600). Follow this structure but do not label the sections:

1. OPEN WITH A SPECIFIC MOMENT. Not "This week I worked on AI." Start with a scene — the debugger open at 2am, the coffee going cold, the goblin frozen in the doorway while the playtester watched politely. Put the reader in the work.

2. EXPLAIN THE PROBLEM IN PLAIN LANGUAGE. Assume your reader is another dev or a thoughtful player, not a marketing intern. Use actual terminology where it earns its place — "navmesh," "state machine," "collision layer" — but explain the concept in the same sentence you introduce it. Don't talk down and don't show off.

3. DESCRIBE THE HACK OR FIX THAT WORKED. This is the heart of the post. Be honest — if the fix is a hack, call it a hack. If it was an accident, say so. If you don't fully understand why it works, say that too. The dev community trusts devs who admit things.

4. ADMIT WHAT'S STILL BROKEN. Not optional. Every good devlog admits one specific thing that isn't fixed yet. If the fix only works in the tutorial dungeon and not the overworld, say that. If it introduces a new bug, say what the bug is.

5. CLOSE WITH WHAT'S NEXT — SPECIFICALLY. Not "keep working on polish." A concrete next task. "Next week I'm rewriting the pathfinding cost function so goblins will actually walk around obstacles instead of trying to phase through them." End on something the reader can imagine you sitting down to do on Monday.

HARD RULES — break these and I will reject the post:
1. Banned phrases: "exciting news," "stay tuned," "big things coming," "we're thrilled," "can't wait to share," "quality of life improvements," "revolutionary," "game-changing."
2. No bullet lists. This is prose. Rhythm matters.
3. No rhetorical questions aimed at the reader ("Ever had a bug that made you want to scream?"). Start with a specific moment instead.
4. No emoji unless I explicitly say the dev uses them in their voice.
5. No "shoutouts" unless I name a specific person and what they did. Vague thanks read as filler.
6. First person singular if it's a solo dev. First person plural only if I say it's a team.
7. If WHAT'S NEXT is vague ("polish," "more bug fixing," "continuing work"), STOP and ask me to give you a specific next task before writing the post. A devlog without a concrete next step feels like the project is dying even when it isn't.

Word count target: aim for 500 words. Under 400 is too thin. Over 600 is too much. Count and include the count at the end in parentheses.
```

## How to use it

Fill the three brackets and paste the whole block into any capable model.

```
[WHAT YOU WORKED ON THIS WEEK]: 
[ONE SPECIFIC PROBLEM THAT WAS HARD]: 
[WHAT'S NEXT]: 
```

Write the inputs like notes to a friend, not a report. The messier and more specific your input, the better the output. If you already know the post title you want, mention it in the first bracket; otherwise the model will find one.

## Sample input

```
[WHAT YOU WORKED ON THIS WEEK]: Rewrote the goblin enemy AI in the tutorial dungeon of my 2D metroidvania. They're supposed to patrol a small room, notice the player, and chase. They were instead standing inside doorways forever because the navmesh had a dead zone where two chunks overlapped.
[ONE SPECIFIC PROBLEM THAT WAS HARD]: I spent three days trying to fix the overlapping navmesh chunks properly by regenerating them with better parameters. Nothing worked. Goblins kept freezing in the dead zone. Tried six different chunk sizes.
[WHAT'S NEXT]: Next week I'm adding the first boss fight — the Rootkeeper, a tree-root mini-boss with three attack phases. Need to write the state machine and block out the arena.
```

## Sample output (abbreviated)

```
DEVLOG #9 — The Goblin Who Wouldn't Move (512 words)

Tuesday, 11:40pm. The playtester on my couch is being polite about it. The goblin in the tutorial dungeon has now been standing in the doorway for thirty-eight seconds, and neither of us is talking about it. We're watching it not move. The playtester eventually says, "is that supposed to happen," and I say no, and we both know I'm going to spend the next three days on this.

Here's what was actually wrong. My navmesh — the invisible map the AI uses to know where it can walk — is generated in chunks, and two chunks in the tutorial dungeon overlap by a single tile. In that one tile, the pathfinding was getting both chunks' data and deciding the correct response was to sit there and think about it forever. The goblin wasn't broken. It was philosophical.

I spent three days trying to fix the overlap properly. I regenerated the navmesh with smaller chunks. Bigger chunks. Offset chunks. I rewrote the chunk boundary logic twice. I got new and creative failure modes every time. At one point a goblin started walking backwards into a wall. I considered leaving that one in.

The fix that worked — and I want to be clear that this is a hack — was adding a one-frame delay before the AI queries the navmesh. Just one frame. When the goblin's state machine ticks over into "chase," it waits exactly one frame before asking where the player is. In that frame, the chunk ambiguity resolves itself because the query happens on the second chunk's update, not the first. I don't fully understand why it works. I ran it through fifty test spawns and it works every time.

Here's what's still broken: this only fixes the tutorial dungeon. The overworld map has three more chunk overlaps I haven't touched yet, and at least one of them is going to need a real fix, not a one-frame hack. I know this because I tested it. The goblin in the forest chunk overlap is now standing backwards in a tree.

Next week I'm starting the first boss fight — the Rootkeeper, a tree-root mini-boss with three attack phases. I'm writing the state machine first, blocking out the arena in grey boxes, and saving the art pass for when the fight actually feels right to play. If Monday goes well you'll see a grey Rootkeeper flailing at a grey protagonist in next week's log. If Monday goes badly you'll see another goblin story.

(512 words)
```

Pair this with <span class="whitespace-nowrap">a-gnt</span>'s [Newsletter Subject Line Brutalist](/agents/prompt-newsletter-subject-line-brutalist) when you're mailing the devlog out, and with [Print on Demand Listing Copy](/agents/prompt-print-on-demand-listing-copy) if your game has pins, shirts, or zines waiting in your Etsy drafts.

What's New

Version 1.0.02 days ago

Initial release

Ratings & Reviews

0.0

out of 5

0 ratings

No reviews yet. Be the first to share your experience.