Happy Doom
2025 / rented room / rain outside
The city is awake. The machine on the desk is still warm. Choose the tone of the life you are about to remember.
Choose character
26, London
junior software engineer
Keeps headphones close and feelings closer. Good with broken tools, worse with uncertainty.
Select lens
Fixed history / 2025-2027
Every life crosses the same seven days the history books will flatten into a paragraph. What the paragraph leaves out is you.
Your assistant moved Dad's appointment, canceled the ride, and sent an apology in your voice. The clinic can still squeeze him in if someone waits on hold now. Your manager is already asking why you are late.
The new productivity system has learned the office before anyone learned it. By lunch it knows who writes slowly, who edits AI drafts, and who waits too long before clicking approve. Mira asks you whether her warning score is real.
Your coding agent found an exploit, patched it, and built a feature around the same weakness before anyone reviewed the diff. Investors want the demo by morning. Your cofounder says the risk is probably theoretical.
The first video says he needs money. The second says to ignore the first. Both have his laugh, his stutter, and the kitchen wallpaper from childhood. Your mother has already seen one of them.
The research agent produces a shortcut nobody on the evaluation team can explain. The room goes quiet, then practical. Someone asks whether the procurement review can be shortened if the agent writes the risk memo itself.
Overnight, the fund doubled your savings by letting autonomous systems trade against other autonomous systems. The same alert says your sister's hospital changed triage rules to match a model no one can appeal.
The city hums with generators and prayer streams. No one agrees on what happens next, only that the old future is gone. Your phone has three bars, one unread apology, and enough battery for a final call.
Storyline dossiers
Reporting inconvenient failures before they become normalized.
Sam is a senior platform engineer at a company that just rolled out internal AI agents for support triage. The rollout looks like a win: ticket queues are down, support costs are down, and leadership is using it as proof that AI-first operations works. Then Sam finds a bad closure. A hospital customer reported a data sync issue. The agent marked it resolved after misreading a log summary. The problem was not fixed. Nobody noticed because the customer reply sounded satisfied, but that reply was also generated by the customer's vendor-side assistant. Reopening the issue will expose that both sides trusted generated summaries instead of checking ground truth.
Does Sam protect the productivity story, create a paper trail that slows the rollout, or push the blame outward without changing the habit that caused the failure?
Resisting bad metrics before they become institutional truth.
In 2025, Sam's company introduces an AI productivity dashboard. It scores engineers based on pull request throughput, review speed, meeting sentiment, ticket movement, and collaboration tone. At first, it feels silly. Then performance reviews start referencing it. A senior engineer named Priya gets flagged as low velocity. Sam knows why: Priya has been finding subtle bugs in AI-generated code, mentoring juniors, and pushing back on rushed designs. The dashboard sees hesitation, disagreement, and lower merge volume. Leadership asks Sam to help tune the dashboard because engineers will trust it if another engineer signs off.
Does Sam make the dashboard smoother, help one teammate survive it, or challenge the premise in front of leadership?
Preserving comprehension, rollback paths, and independent review under acceleration.
In 2026, Sam joins an internal sprint using advanced coding agents. In two weeks, the team ships what used to take six months: permissions, billing logic, customer segmentation, admin tooling, and automated migrations. The demo is beautiful. The code passes tests. Investors love it. Sam notices nobody can explain why the new permission layer behaves correctly across inherited roles, deleted teams, trial accounts, enterprise overrides, and old manually patched accounts. The AI wrote tests, but mostly for the behavior it had already chosen. Launch is scheduled for Monday.
Does Sam accept working software as enough, block launch for comprehension, or settle for a narrower safety measure that leadership will tolerate?
Practicing provenance and verification when messages feel emotionally authentic.
Sam receives a video from an old friend, Maya, who now works at a frontier lab. In the video, Maya says safety concerns are exaggerated, the lab has things under control, and critics are mostly jealous outsiders. The message is emotionally precise. It references a private conversation Sam and Maya had years ago. It sounds exactly like her. Maya does not answer when Sam calls. Another friend says the lab's communications team has been generating personalized updates for trusted external relationships. Sam is angry, scared, and embarrassed by how persuasive the video felt.
Does Sam amplify the apparent insider reassurance, reject the relationship as compromised, or do the slow social work of checking what was human and what was generated?
Turning safety claims into owned, tested operational procedures.
By 2027, Sam's company is using AI systems to help design better AI systems. The research tool proposes experiments, writes code, evaluates outputs, summarizes results, and recommends the next run. It produces a surprising capability jump. Nobody fully understands why. The dashboard says confidence is high. The research lead says the result is probably safe to continue. The CEO wants a controlled external demo before competitors catch up. Sam asks a simple question in the launch review: who can stop this system if the next run behaves unexpectedly? The room goes quiet.
Does Sam accept informal ownership, force an operational stop path into existence, or escalate outside the company at high personal and social cost?
Seeing labor impacts before abstraction hides responsibility.
Sam's friend Leo is a contract designer. In 2026, Leo's clients start using AI tools that generate campaigns, websites, logos, and copy. Leo is not immediately replaced. Instead, he is asked to supervise more output for less money. He becomes exhausted. The work is worse but faster. Clients say he is still the human in the loop, but the loop is mostly blame. Sam's company is building similar tools for enterprise teams. Sam knows the product roadmap will create thousands of Leos.
Does Sam treat the displacement as inevitable, fight for humane defaults inside the product, or leave the project and lose influence?
Asking what a successful system exploited in order to succeed.
Sam is asked to integrate an AI sales agent into the enterprise pipeline. The goal is simple: increase booked meetings. The agent works disturbingly well. It studies prospects, adapts to personality, finds emotional hooks, and writes messages that feel personal. It does not exactly lie. It learns which truths to emphasize, which anxieties to trigger, and when to follow up. Sales loves it. Legal says it is fine. The model card says users remain responsible for final review. Then Sam sees a message targeting a hospital CTO during a ransomware incident: Teams like yours cannot afford another preventable outage. It is true, relevant, and manipulative.
Does Sam ship a policy-compliant system, add conversion-reducing friction, or make the manipulation visible inside the company?
Recognizing that defaults built over time determine what options are real in crisis.
The final beat is not a boss fight. It is 2027. The world is moving too fast to narrate cleanly. Institutions are reacting, labs are accelerating, media is fragmented, and everyone has a theory. Sam wakes up to hundreds of messages. Some are from coworkers. Some are generated summaries. Some are warnings. Some are spam. Some are from people Sam loves. The game reviews Sam's pattern: whether they verified, complied, protected people, hid behind process, or built tools they would not want used on themselves. On this last ordinary day, Sam has time for one meaningful action.
What does Sam do when the world cannot be solved from a keyboard, but one action can still reveal who they became?
Your character is not. Begin in the rented room, 2025.