This is a submission for the Sanity Challenge, Path One: Ship an Agent That Queries Real Content
What I Built
I built TrueSpecs, a game compatibility checker powered by an AI agent that reasons over structured community data.
Before I explain what it does, let me tell you why I built it.
I've been gaming since I was a kid. On my old laptop, I could run GTA Vice City fine. It wasn't a powerful machine. It had less RAM than my current laptop. But it ran. Then I got a newer laptop. More RAM. Better specs. I installed Vice City again, excited to play. It crashed on launch. Same game. Better hardware. It wouldn't run.
That was the moment I realized official system requirements are almost useless.
They tell you the minimum. They don't tell you that the game needs DirectPlay, which Windows 11 diasabled by default. They don't tell you that it forces a 640x480 resolution that modern drivers reject. They don't tell you that a game with an RTX 3070 listed as recommended will still stutter on an RTX 5090 because the stutter is built into the engine.
So what do you do? You go to Google. You search why does this game crash on my PC. You find a Steam review from 2022. You find a Reddit thread from 2023. You find a ProtonDB report from six months ago. Each one tells you something different. Some say it runs fine. Some say it doesn't. Some say you need to add dx12 to launch options. Some say you need to update your drivers. Some say the anti cheat blocks Linux entirely and there's nothing you can do.
You spend hours digging through forum posts. You install the game. It crashes. You refund it. You try another game. Same cycle.
Games are expensive. If you buy something and it doesn't run, that money is gone. Refunds exist but they're not guaranteed and the process is slow.
This is the pain TrueSpecs solves.
What It Does
TrueSpecs is an agent that answers one question. Will this game run on my machine?
You give it your specs. It queries a structured knowledge base of official requirements and real user reports. It finds every report from users with similar hardware. It finds the conflicts where official specs say one thing and real users say another. It surfaces known fixes.
And then it gives you an answer. Not a guess. Not a marketing line. An evidence backed answer with sources.
If it doesn't have data on the game you're asking about, it says so. It doesn't hallucinate. It doesn't pretend.
The Data
I curated data for four games.
GTA Vice City Definitive Edition
Battlefield 6
007 First Light
Star Wars Outlaws
For each game I collected official system requirements from the Steam store page, real user reports from Steam reviews and ProtonDB, and known fixes from PCGamingWiki.
That's 44 user reports across 4 games. Each one has the user's actual hardware specs, platform, result, and notes.
Demo
Code
TrueSpecs
A game compatibility agent that answers "will this run on my machine?" using structured community data, not just official specs.
The Problem
Official system requirements are often misleading. A game can list an RTX 2060 as minimum, but real users with RTX 5090s still report stutters "built into the core." A game can say "Windows 10" but block Linux entirely via anti-cheat. Spec sheets are necessary, but not sufficient.
What TrueSpecs Does
TrueSpecs queries a structured knowledge base of official specs and real user reports, then answers compatibility questions with evidence. When sources conflict, it shows both sides. It never guesses.
How It Works
-
Content model (Sanity): Four content types
-
game— official requirements, ProtonDB rating, known fixes -
userReport— real user specs, platform, result, notes -
source— where each claim came from (with trust score)
-
- Query layer: GROQ queries traverse the relationships (game → userReport → source).
- Agent:…
How I Used Sanity
I modeled three content types.
game for official requirements, ProtonDB rating, Steam Deck verified status, and known fixes
userReport for real user specs like OS, RAM, GPU, CPU, platform, result, notes, and source URL
source for where each claim came from with a trust score
I deployed the schema and created a Sanity Context MCP endpoint. The agent connects to that endpoint and receives tools for querying the structured content.
When a user asks a compatibility question, the agent traverses the relationships. game to userReport to source. It pulls the official requirements and then finds all user reports linked to that game. It compares the user's specs against real reports.
The Battlefield 6 example is the clearest. A user with an RTX 3070 and 32GB RAM asks if the game will run. The official specs say yes. The RTX 3070 is above the recommended RTX 3060 Ti. But the agent also finds a user report from someone with the exact same RTX 3070 who was locked out by anti cheat. The game installed then refused to launch.
The agent shows both. Official says yes. Reality says no. It doesn't pick a winner. It shows the evidence and lets you decide.
That's the content model doing the work. Without the relationships between games, reports, and sources, the agent couldn't surface that conflict. A keyword search would find the official specs. It wouldn't find the RTX 3070 user who got locked out. The structure is what makes the answer possible.
Sanity Project Details
bc0hielv
Top comments (0)