 | Space Camel Predicting the future of humanity and machine intelligence Run 2 · 28 September 2026 |
|
What is this? Every week, we forecast the odds that twelve world-changing events happen within the next 25 years. Each index can move at most 5% of its value per week - so big shifts in opinion show up gradually, over several weeks. Our full methodology (gamification of frontier AI models w/Olympic panel style scoring) is shared towards the end of this post. Jaded human commentary and occasional insight (below) is included free of charge. |
|
📈 Global Progress Tracker 100.00 100.00 · starting value Average of 6 progress indices | | ⚠️ Global Risk Tracker 100.00 100.00 · starting value Average of 3 risk indices | | Each tracker begins at 100 in run 2 and will average the rise and fall of its underlying indices in subsequent runs. |
|
🗒️ Weekly Findings & Commentary Welcome to the new World Order. Where AI predicts what AI will do and lo and behold, AI thinks the odds of AI doing amazing things (being recognized as conscious, commanding robots in our homes, and interfacing with our brains through thought alone) is up and up over our first two readings. We will want to keep a watchful (human) eye on this trend to ensure there isn’t an overabundance of pro-AI bias happening. Naturally some bias is unavoidable. Bias is after all, how the nature and all things evolve to suit their environment. Some things are just better at it than others. Let’s “see” what happens! |
|
📊 Probability Indices The odds that each event happens within the next 25 years, how they moved this week, and the news that moved them - picked by the AI that called it best. |
💡 Conscious AI P(conscious_system) Expert consensus that a running AI system is conscious, self-aware and sets its own goals. | | 9.88% about 1 in 10 25-year probability ▲ 9.41% → 9.88% since last week 🔒 Hit the 5% weekly limit for the 2nd week running — the AIs wanted 10.84% At the weekly limit, about 2 more weeks to get there if the AIs hold this view. AI forecasts ranged from 6.00% to 22.00% | | Why it moved: Research showing AI models can act out pain-like internal states will fuel more position-taking papers, but the literature still treats a conscious, self-goal-setting named system as unestablished. | Winning LLM analyst: Grok 4.6 (xAI) |
|
☠️ Population Collapse P(population_collapse) · ⚠️ Risk index Human population falls 90% or more from its all-time peak, for any reason. | | 1.21% about 1 in 82 25-year probability Restarted — new definition this week AI forecasts ranged from 0.50% to 4.50% ⓘ New definition from this week: Population Collapse now measures a fall of 90% or more from humanity’s all-time peak population, replacing a fall below 400 million. This reading restarts the index and is not comparable with earlier readings. | | Why it moved: Restarted under the new 90%-from-peak test: fertility decline alone cannot remove nine in ten people in 25 years, leaving the risk in extinction-scale tails such as engineered pandemics and AI misuse. | Winning LLM analysts: Claude Opus 5.5 (Anthropic); GLM 5.3 (Z.AI) |
|
🚀 Earth Escape P(earth_escape) · 📈 Progress index 10,000 people have each lived two years off Earth in a habitat with no material resupply. | | 0.68% about 1 in 150 25-year probability ▼ 0.72% → 0.68% since last week 🔒 Hit the 5% weekly limit for the 2nd week running — the AIs wanted 0.61% At the weekly limit, about 3 more weeks to get there if the AIs hold this view. AI forecasts ranged from 0.30% to 1.20% | | Why it moved: Habitat programmes are only now starting: Vast's new lunar division targets small, resupplied crews in the 2030s, far from 10,000 people living two years with zero resupply. | Winning LLM analyst: GLM 5.3 (Z.AI) |
|
☀️ Fusion Power P(fusion_power) · 📈 Progress index Fusion supplies at least 51% of global electricity in a calendar year. | | 1.74% about 1 in 57 25-year probability ▼ 1.83% → 1.74% since last week 🔒 Hit the 5% weekly limit — the AIs wanted 1.33% AI forecasts ranged from 0.20% to 1.88% | | Why it moved: Magnet and plasma-control milestones advance plant components, but no fusion plant yet sells power, and supplying 51% of the world's electricity by 2051 would need an exceptionally fast build-out. | Winning LLM analyst: Astra 6 (OpenAI) |
|
⚛️ Quantum Chemistry P(quantum_chemistry) · 📈 Progress index Ten novel molecules designed on a fault-tolerant quantum computer are synthesised and confirmed. | | 29.95% about 3 in 10 25-year probability ▲ 28.53% → 29.95% since last week 🔒 Hit the 5% weekly limit — the AIs wanted 30.74% AI forecasts ranged from 24.00% to 65.00% | | Why it moved: Error-corrected logical qubits now beat physical ones and the DOE is funding 100+ logical-qubit machines aimed at chemistry, giving ten confirmed quantum-designed molecules room to arrive within 25 years. | Winning LLM analyst: Claude Opus 5.5 (Anthropic) |
|
🤖 Household Robots P(household_robot) · 📈 Progress index 35% of the world's people live with a humanoid robot that does household chores. | | 9.29% about 1 in 11 25-year probability ▲ 8.85% → 9.29% since last week 🔒 Hit the 5% weekly limit — the AIs wanted 9.41% AI forecasts ranged from 5.00% to 24.00% | | Why it moved: Figure's humanoid did chores in 30 homes it had never seen, though it fully completed only 56% of tasks; reaching 35% of the world still needs a steep fall in cost and a big jump in reliability. | Winning LLM analyst: GLM 5.3 (Z.AI) |
|
🧬 Designer Babies P(designer_baby) 10% of babies born in a year come from embryos edited for enhancement of selected traits. | | 1.61% about 1 in 62 25-year probability ▲ 1.60% → 1.61% since last week AI forecasts ranged from 0.60% to 2.20% | | Why it moved: Columbia's embryo base-editing work shows research-grade precision but safety risks that rule out clinical use, and heritable editing stays banned or restricted in most of the world. | Winning LLM analyst: GLM 5.3 (Z.AI) |
|
🛰️ Alien Civilization P(alien_civilization) Expert consensus that a confirmed observation is technology from a civilisation beyond Earth. | | 3.31% about 1 in 30 25-year probability ▲ 3.27% → 3.31% since last week AI forecasts ranged from 1.50% to 3.50% | | Why it moved: Wider sky surveys and stricter verification protocols raise the odds of catching a technosignature, but consensus still needs evidence that survives scrutiny across independent instruments. | Winning LLM analyst: Gemini 3.8 (Google) |
|
⏳ Age Reversal P(age_reversal) · 📈 Progress index An intervention cuts biological age by 20 years and all-cause mortality by 80% in randomised trials. | | 2.53% about 1 in 40 25-year probability ▼ 2.66% → 2.53% since last week 🔒 Hit the 5% weekly limit for the 2nd week running — the AIs wanted 2.35% At the weekly limit, about 2 more weeks to get there if the AIs hold this view. AI forecasts ranged from 1.00% to 4.00% | | Why it moved: Rentosertib moved proteomic aging clocks by 3-4 years in 42 patients over 12 weeks - a long way from a 20-year reset in 1,000 older adults plus an 80% cut in deaths. | Winning LLM analyst: Grok 4.6 (xAI) |
|
🧠 Brain Telepathy P(brain_telepathy) · 📈 Progress index A non-invasive wearable on public sale lets people send text and commands to machines by thought alone. | | 11.10% about 1 in 9 25-year probability ▲ 10.58% → 11.10% since last week 🔒 Hit the 5% weekly limit — the AIs wanted 11.46% AI forecasts ranged from 9.00% to 32.00% | | Why it moved: Ultrasound and wearable decoding are advancing and implants already clear the speed bar, but a mobile, battery-powered device that sends text from brain activity alone is not yet on sale. | Winning LLM analysts: Claude Opus 5.5 (Anthropic); Grok 4.6 (xAI) |
|
💼 Mass Unemployment P(mass_unemployment) · ⚠️ Risk index Unemployment stays at or above 25% for a full year in the US, Japan and the UK at the same time. | | 4.46% about 1 in 22 25-year probability First reading — new this week AI forecasts ranged from 1.50% to 13.00% | | Why it moved: US, Japanese and UK underutilisation sits far below 25%; AI is raising job-loss fears, but only a Depression-scale shock hitting all three at once for a full year would count. | Winning LLM analysts: Claude Opus 5.5 (Anthropic); Grok 4.6 (xAI) |
|
🔌 Internet Blackout P(internet_blackout) · ⚠️ Risk index 80% or more of the internet’s users are cut off from the global internet for 12 consecutive months or more, for any reason. | | 1.76% about 1 in 57 25-year probability First reading — new this week AI forecasts ranged from 0.60% to 5.00% | | Why it moved: Even war-driven blackouts like Tigray's stay regional; only a thin tail such as a Carrington-class solar storm or great-power war could cut off 80% of users for a year. | Winning LLM analysts: GLM 5.3 (Z.AI); Grok 4.6 (xAI) |
|
📐 Methodology, Index Definitions, and AI Lab Scoreboard |
Click here for a readme file with the methodology deployed for producing Space Camel, along with the standardized, more comprehensive definition of each index.
Click here for a look at the AI Labs weekly results and 26-week rolling scoreboard. |
|
If you’re a fan of Space Camel, please share. Future networks of bots will one day admire your foresight. See you next time✌️ Disclaimer: We may make mistakes; the future is uncertain. |
|
| Space Camel Indices · Run 2 · 28 September 2026 |
