Most demos show a tool at its best. This isn't that. Below is the real GM Brief that Aftermath Engine generated from a real session of mine, shown in full, with my honest read on what it got right, what it got wrong, and where I had to step in. If you're skeptical about whether an AI can actually track your campaign without making things up, this is the page I'd want you to read.
The session is from my own homebrew Electric Bastionland game, run live on TWK Live by me, GM Jeremy. I'm using it rather than a published setting because it's entirely my own material to share, and because Electric Bastionland is rules-light and almost entirely improvised, which is a harder test for the tool than a structured combat-heavy game. If it can track a loose, weird, conversation-driven session, the structured ones are easy by comparison.
Don't take my word for any of this. The entire session is public on our stream, so you can watch the raw session yourself and check the brief against it line by line. The link is at the bottom. That's the whole point: here's what the tool produced, here's the unedited session it came from, go verify me.
The crew inherited a 4,017-pound debt from a dead reverend and were sent into the fog-choked Off-Grey District to recover the Bell of Black Fog from a ruined chapel. Along the way they met their own mirror-doubles in the fog, a weeping singer who bottled fog in glass jars, a bureaucratic "Self-Finders Union" demanding to verify their identities, and a Bell Broker who traded small bells for spoken confessions. They fought a group of masked attackers of uncertain allegiance, took the disguises off the bodies and wore them into the chapel, and finally reached the nave, ending the session standing inside as disembodied voices spoke their names back to them.
That's the raw material. Here's exactly what the tool produced from it.
This is the unedited output, copied straight from Aftermath Engine. Nothing added, nothing cleaned up.
The crew inherited a 4,017-pound debt from the late Reverend Tarquin Sill and ventured into the fog-choked Off-Grey District to recover the Bell of Black Fog from Fogway Chapel. After encountering mirror-doubles, a weeping singer with bottled fog, the Self-Finders Union, and the mysterious Bell Broker, they reached the chapel, now standing inside the nave as disembodied voices speak their names back to them.
The Bell Broker: Neutral to cryptically helpful. Accepted confessions, provided small bells, and offered the only concrete guidance: "The bell you seek is near. You're halfway there."
The Weeping Singer (Choir of One): Terrified of whoever held the glass bell (Mater). Gave them a fog jar and begged them to leave. Unresolved fear, may reappear or send word.
| Faction | Disposition | Toward Party | Key NPC | Active Goal | Pressure |
|---|---|---|---|---|---|
| Office of Spiritual Burdens | Bureaucratic | Creditor | Mr. Porridge | Recover the Bell of Black Fog | Debt accrues hourly |
| Mayor Impanzi's Administration | Transactional | Creditor | Mayor Impanzi (mock chimp) | Collect 10,000-pound debt | Political leverage |
| Self-Finders Union | Suspicious | Monitoring | Unnamed clerks | Identify and document "irregular duplications" | May verify identities later |
| Fog Saints | Unknown | Unknown | None yet encountered alive | Unclear, worship or guard the fog? | Party wearing their garb |
| The Bell Broker | Neutral | Helpful (for confessions) | The Bell Broker | Trades bells for confessions | Knows more than he shared |
| The Fog itself | Hostile/archival | Predatory | N/A | Preserves and replicates memories | Doubles, whispers, psychological pressure |
First session on record.
That's the complete brief. Now my honest read on it.
This is the part that matters most, so I'll be specific. The brief did not invent events. Every major beat appears in it, correctly. More impressively, it captured details I'd have expected an AI to fumble.
The three confessions to the Bell Broker were recorded exactly, each tied to the right character: Lloyd's confession involving a family member, Mater's admission of a past cruelty, and Sandoval's confession to a killing. (I've kept the specifics vague here out of respect for my players; they're in the raw session if you really want them.) These are throwaway improvised lines buried deep in a rambling session, and the tool attached each one to the correct character.
It correctly captured the Bell Broker's one piece of concrete guidance, that the bell was "near" and the party was "halfway there." It tracked the skull-faced woman from Sandoval's vision and her line about "choosing differently." It noted the footprints in the fog matching Mater's running shoes, the splinter Sandoval took from the chapel door, and that the session ended with a voice speaking the party's names inside the nave.
For a tool whose entire job is to not hallucinate your campaign, this is the evidence I'd point to. It tracked the things that actually happened and connected them to the right people.
Now the honest part. It made mistakes, and I'd rather you see them here than discover them yourself.
The clearest error is right there in the Consequence Register: it says "Sandoval, Larry, and one other" put on the Fog Saint disguises. The session is unambiguous, the three who donned disguises were Sandoval, Larry the lackey, and Mater. The tool got two of the three right and vaguely hedged the third as "one other" when the session names him plainly. That's a real miss, and the kind of thing I correct with the inline editor before relying on a brief.
A subtler one: the brief treats the masked attackers as confirmed "Fog Saint cultists" in the Pressure Points. In the actual session they were just masked young men of uncertain affiliation, the party guessed they might be Fog Saints, but it was never confirmed. The tool firmed up an ambiguous encounter into a more definite faction read than the session supported. Not invented, but more confident than the evidence warranted.
It also made a couple of reasonable-but-unstated inferences in the faction table, like tying Mr. Porridge to the Office of Spiritual Burdens specifically, when that connection was implied rather than said. Defensible calls, but calls the tool made, not facts stated at the table.
There were minor name wobbles too. One character's name got said several different ways during play, and the tool quietly standardized it. Probably the right call, but worth knowing it makes that kind of judgment.
The brief skipped the texture: the comedy of the party being too broke to afford anything at a bar, the running gag of finding cursed dice, the streaming hiccups at the top of the session. That's correct behavior, that's noise, not campaign state, but it's worth knowing the tool filters aggressively for what changes your world versus what was just fun at the table. If you want a vibes recap, this isn't that. It's a working brief.
On this session, the tool got the events and consequences right with no fabrication, made one clear factual error I had to fix, and made a few inferences that were defensible but firmer than I'd have written myself. The fixes took me a couple of minutes with the inline editor.
Is it perfect? No. Is it worth it to me? Yes, easily. Reconstructing that session by hand from memory or scrubbing back through the recording would have cost me an hour or more, and I'd have missed things, I always do. The tool got me most of the way there in a few minutes, and the corrections were fast and obvious because I know my own game.
That's the deal I'd make every week, and do. If you'd rather build your session notes by hand and enjoy it, you genuinely might not need this. If your post-session prep is a chore you'd pay to shrink, this is what it actually looks like, flaws included.
Don't trust the brief? Good, you shouldn't trust any AI output blind. Watch the full session yourself and check every claim above against what actually happened at the table:
And if you want to see what your own last session looks like through Aftermath Engine, you can try it free.