# AI futures: from AI to RSI to superintelligence | AGI Explore conditional AI futures, societal changes and business opportunities. Change assumptions, inspect evidence, and keep a reproducible research record. --- Canonical page: https://agiscorecard.com/ai-futures Dated company sources: https://agiscorecard.com/futures-assets/snapshot.json Conditional model baseline: https://agiscorecard.com/futures-assets/forecast.json (not calibrated probabilities) This Markdown mirror is generated from the page; the HTML page is canonical. ## Conditional AI future simulation AGI RESEARCH LAB / 01 The future is a range. Explore your assumptions. Explore conditional AI futures, societal changes and business opportunities. Change assumptions, inspect evidence, and keep a reproducible research record. Build my scenario ↗ (#simulator)Understand the method → (#method) Free to explore · No account required · Runs in your browser 01 / NOW AI Intelligent tools → 02 / IF RSI Recursive improvement → 03 / IF SI Superintelligence A research framework. No stage or arrival date is guaranteed. Simulator (#simulator)Companies & signals (#companies)Society (#society)Business & markets (#opportunities)My research (#my-record)Method (#method) 01 / SCENARIO EXPLORER What changes when your assumptions change? 1,024 simulated paths These sliders are illustrative assumptions, not measurements of the world. Compare a baseline, then test the conditions you think are plausible. My assumptions Compute investment 55 Capital and effective training / inference resources. Research automation 50 How strongly AI helps develop the next generation. Energy & delivery 50 Power, chips and physical infrastructure availability. Deployment constraints 50 Higher values slow deployment in this simplified model. Explore through 2030 2035 2040 2045 Reproduce this run Random seed Run my scenario → Reset to baseline Same model, evidence and seed → same result. Illustrative baseline · 2035 Loading model… RSI proxy crossed — Share of simulated paths SI proxy crossed — Share of simulated paths RSI not reached — Within selected horizon Illustrative capability index Median · shaded 10th–90th percentiles Conditional simulation shares, not calibrated probabilities. The index is a model variable, not an AGI benchmark. Keep this scenario & continue researching → (#my-record) Which assumptions move the result? One parameter +10 points (−10 above 90), same seed, 384 paths. This is model sensitivity, not causation measured in the world. 02 / EVIDENCE RADAR Follow the companies. Question the signal. Check latest snapshot ↻ Last snapshot check: 2026-10-31T16:20:44Z The existing daily publishing job checks five official sources. This page checks the published snapshot every five minutes. Feed metadata is automatic discovery, not independent verification. 01 OpenAI Professional agents, model economics and distribution. Watch paid retention, task reliability and actual cost per completed task. Official source ↗ (https://openai.com/news/) 02 Anthropic Enterprise workflows, coding and controlled autonomy. Watch production deployment, cloud commitments and customer margins. Official source ↗ (https://www.anthropic.com/news) 03 Meta Consumer distribution, open ecosystems and ad monetization. Watch product retention, licensing, capital spending and free cash flow. Official source ↗ (https://about.fb.com/news/) 04 SpaceX / xAI Grok, compute, connectivity and infrastructure integration. Watch disclosed corporate structure, segment economics and actual delivery. Do not infer a tradable ticker from a brand. Official source ↗ (https://x.ai/news) 05 Google DeepMind Research automation, cloud and existing distribution. Watch independent replication, cloud profit and cash returns on infrastructure. Official source ↗ (https://deepmind.google/blog/) Recent source items & feed health Use latest evidence for this scenario 03 / HUMAN CONSEQUENCES Progress changes more than technology. Conditional research questions to revisit as evidence changes. These are editorial scenarios, not outputs trained on future outcomes. 01 Work & income Automation can expand a small team’s reach. Junior tasks and bargaining power may shrink first. Track wages, paid hours and value added—not task benchmarks alone. 02 Science & education Cheaper experiment design and tutoring matter only if replication, clinical outcomes and long-term learning improve. Physical validation can remain the bottleneck. 03 Energy & infrastructure Demand may grow for grids, cooling and interconnects. Efficiency gains, overbuilding and low utilization can reverse the investment case. 04 Power & governance Ownership and access determine who captures productivity gains. Watch concentration, portability, audit quality and whether audits actually change deployments. 05 Security & trust More capable agents can strengthen defense and scale misuse. Track actual incidents, loss severity, access controls and emergency intervention capacity. 06 Life beyond work Under safe and broadly shared progress, services could become cheaper. Tax systems, public services and accountability may adapt slowly; abundance is not automatic. 04 / OPPORTUNITY MAP Find a buyer. Test the economics. 01 Reliable industry workflows Sell a measurable result to a named budget owner. Test three paid pilots in six weeks; include correction and delivery cost. 02 Verification & safety Sell audit readiness or lower failure cost. A free evaluation without a buyer or renewal budget is weak demand evidence. 03 Physical research automation Pair models with instruments and reproducible feedback. Compare full experiment cycles, certification and support costs. 04 Energy & compute efficiency Measure realized savings after retrofit, financing and operations. Nominal efficiency is not the same as positive cash flow. Market exposure to investigate Open investment research ↗ (/invest) Examples for research, not a ranked buy list. No live prices or return forecasts. Technical progress can coexist with weak shareholder returns. Confirm listing, financials, valuation and risks in current filings. Theme Examples What can break the thesis Compute supply NVDA · AMD · AVGO · TSM · ASML Orders, margins, customer concentration, export rules and custom-chip substitution. Cloud & distribution MSFT · AMZN · GOOGL · META Incremental AI revenue versus depreciation, capital spending and price competition. Power & cooling ETN · VRT · SU.PA Grid delivery, backlog conversion, utilization and cash collection. China platforms 0700.HK · 9988.HK / BABA Paid adoption, cloud margins, regulatory changes and financing dilution. 05 / MAKE IT YOURS Keep the reasoning. Revisit the decision. What would change my mind? Save on this device Export record Import record Clear local copy Local saves and exports are free. Parameters and notes stay in your browser unless you explicitly save them on the member page. Free account Turn a scenario into a research question. Use the shared AGI account to continue in Jarvis or join discussions. Registration does not include paid cloud storage. Register / sign in ↗ (/auth/account?next=%2Fjarvis&lang=en)Explore Jarvis first → (/jarvis) WHEN YOUR RESEARCH NEEDS CONTINUITY A workspace for your evolving judgment. Existing AGI cloud membership saves this scenario across devices with version history. It uses a separate member key; it is not automatically linked to your free account. 50 workspaces · 10 versions each · 64 KiB per version · 5 MiB total. No auto-renewal or additional AI quota. 9 USDT / 30 days Plus order-matching decimal and network fees. USDT only. Take my scenario to membership ↗ Checking current availability… Opening the page does not upload, place an order or charge you. Save explicitly there. 06 / OPEN THE MODEL A model you can inspect. 01 · System dynamics Quarterly capability and research-autonomy states interact with resources, deployment constraints and feedback. Coefficients are explicit assumptions. 02 · Monte Carlo ensemble 1,024 reproducible paths sample slow, middle and fast regimes with priors 28%, 50%, 22%. The band describes variation inside this model, not all uncertainty. 03 · Evidence weighting Dated official items from the past 90 days can make small heuristic updates. At most one signal per company; announcements are not independent measurements. This is uncalibrated Bayesian-style weighting. 04 · Neural surrogate A 4→12→8→1 neural network learns 200 synthetic simulator examples at the 2035 horizon (160 training, 40 held out). It approximates this simulator, not the real future. It does not replace the main result. Train & inspect locally Exact thresholds, assumptions & limitations RSI proxy: capability ≥ 0.72, autonomy ≥ 0.55 and feedback multiplier ≥ 1.55 for four quarters. SI proxy: after RSI, capability ≥ 0.92 and autonomy ≥ 0.80 for four quarters. These definitions are modeling choices, not agreed scientific standards. The model starts at 2026.75 with capability 0.32 and autonomy 0.13. Coefficients have not been fitted to a historical outcome set. There are no verified RSI/SI outcome labels for calibration. It does not estimate unemployment, geopolitical incidents, company returns or existential risk. Governance only models deployment speed here, not the benefits of safety. Inspect core model source (/futures-assets/engine.mjs) · Inspect source snapshot (/futures-assets/snapshot.json) · Daily computed baseline (/futures-assets/forecast.json) Each daily publication recomputes the default scenario from its source snapshot. Your saved assumptions and old snapshots are preserved until you explicitly change them. Embed this free simulator Copy embed code Company research, social scenarios and the method are readable without JavaScript. Enable it to run a simulation or save your record.