howclose.to

An honest tech tree for civilization · 20 bottleneck technologies

howclose.to

How close are we? - tracked by milestone, metric and setback. Never a fake percentage.

EST. 2026 · UPDATED JULY 2026

The example · Nuclear Fusion · Can we bottle a star?How close are we to fusion power?

19201920Eddington proposes stellar fusion1934Fusion achieved in a lab1958The tokamak is born1968T-3 stuns the field1983JET switches on1991First controlled fusion power1997JET record: Q = 0.672006ITER agreement signed2021Private fusion waveDEC 2022NIF achieves ignitionFEB 2024 - HERENIF ignites repeatedly, gain ~2.42027SPARC targets Q > 10~2035ARC - first electricityUNLOCKS →Endless clean powerFresh water anywhereReal climate repair
The example · Nuclear Fusion · Can we bottle a star?How close are we to fusion power?Open the Nuclear Fusion file → howclose.to/nuclear-fusion

The problems, by fieldThe register, by domain

Everything we're trackingThe full register, grouped by domain

20 hard problems, grouped by the fields working on them. Scroll the list — nothing is cut.Grouped for browsing; ordering within each domain is editorial, not a ranking.

5 fields · 20 problems

Intelligence & Machines

5

AGI

AI can already help with many kinds of thinking, but it still makes confident mistakes and struggles to carry a big job through on its own.Frontier models saturate most static benchmarks, yet reliable long-horizon autonomy, calibrated reasoning and a shared definition of general intelligence all remain open - so any single 'percent to AGI' is a guess dressed as a number.

Next: ARC-AGI-2 and held-out expert evals resistant to training-data leakage
EngineeringacceleratingBroad cognitive competenceLearn many different kinds of work

Brain Connectomics

Researchers have mapped a whole fly brain and a tiny piece of mouse brain. The hard part is scaling that detail across an entire mammal without losing or inventing connections.A whole adult fly is mapped at synapse resolution and a mouse cortical fragment now contains more than half a billion detected synapses; the whole mouse remains an estimated three orders of magnitude larger.

Next: Demonstrate an end-to-end acquisition and proofreading pipeline across a whole mouse region that preserves both local synapses and long-range axons.
ScalingacceleratingWhole mouse synaptic connectomeWire-map an entire mouse brain

DNA Data Storage

DNA can hold computer files in an extraordinarily small sample. The missing step is a useful machine that writes a large archive quickly, finds one file and costs less than today's cold storage.Synthetic DNA has recovered a 200 MB random-access archive without bit errors and automated a complete write-to-read cycle for five bytes, but no public benchmark joins useful payload, throughput, retrieval and cost.

Next: An appliance vendor must publish an audited end-to-end dollars-per-TB and bytes-per-second benchmark.
EngineeringsteadyArchive-class cost and throughputWrite it fast enough and cheaply enough

Fault-Tolerant Quantum

Today's quantum computers are too error-prone to trust. Fixing that means going from a handful of good qubits to millions.Error-corrected qubits have crossed below threshold; the open question is scaling from one logical qubit to the millions a useful machine needs.

Next: Run a multi-logical-qubit universal circuit through repeated correction with a lower total failure rate than its physical implementation.
Lab demoacceleratingIntegrated Universal Logical OperationsCompute while correcting

General Robotics

Robots can walk, run and keep their balance beautifully now. The hard part left is hands and judgment: doing varied tasks all day, safely, without a human quietly steering.Legged locomotion is essentially solved, but dexterous manipulation and open-world autonomy are not, so today's factory and home robots run narrow, supervised, and heavily teleoperated.

Next: Independent long-horizon manipulation trials on tools and deformable objects, e.g. pi-0.6 field runs
Lab demoacceleratingDexterous manipulationHandle unfamiliar objects and tools

Earth & Space

5

Cheap Space Launch

Rockets can now land and fly again, which made space far cheaper - but the truly cheap ticket needs a whole rocket you can reuse in days, and nobody has that yet.Reusable first stages have cut the price of orbit by more than 90% since the Shuttle, but the sub-$100/kg goal depends on rapid full reuse that Starship has not yet demonstrated.

Next: First Starship upper-stage (Ship) tower-catch attempt, targeted for Flight 13/14 in 2026
ScalingacceleratingRapid full reuseReuse the whole rocket, fast

Closed-Loop Life Support

The space station recycles almost all its water, but a self-sustaining habitat still has to close air, food and waste loops at the same time.Water recycling is near exploration grade in orbit, but oxygen, food, waste and maintenance have never been closed together beyond Earth.

Next: An integrated flight demonstration must sustain oxygen recovery above 75% while managing methane and carbon deposits.
EngineeringsteadySeventy-five-percent oxygen recoveryRecover three quarters of the oxygen

Direct Air Capture

Machines can already pull carbon from the sky. The challenge is doing it cheaply - and a million times bigger.Plants that pull CO2 straight from the air are running; the open question is cost per tonne and scaling from thousands to billions of tonnes.

Next: STRATOS must start sequestration and disclose independently verified net output through a full operating year.
ScalingacceleratingHundred-Thousand-Tonne PlantMove beyond small plants

Exoplanet Biosignatures

We know more than 6,000 planets beyond our Solar System, but zero confirmed living ones. The next leap needs a telescope that can separate an Earth-like planet from a star ten billion times brighter.Atmospheric spectroscopy is now routine for some large exoplanets, but no world beyond Earth has yielded a confirmed biosignature; the observatory intended to survey nearby potentially habitable worlds remains in development.

Next: Obtain a repeatable spectrum of an Earth-sized habitable-zone world that constrains several gases and its environmental context.
EngineeringacceleratingDetect biosignatures in planetary contextFind a bundle of life clues

Planetary Defense

We have moved one harmless asteroid. The bigger gap is seeing a dangerous one early enough: more than half of the estimated regional-size population is still unknown.Humanity has demonstrated one asteroid-deflection technique, but the survey needed to find regional-scale threats is only halfway to its statutory goal.

Next: NEO Surveyor launches and begins its five-year infrared survey of bright and dark objects.
EngineeringacceleratingNinety-percent regional-threat surveyFind almost every regional-size threat

Energy

4

Clean Hydrogen

Clean hydrogen could fuel steel, ships, and fertilizer. First it has to get cheap - and get built at scale.Electrolysis works; the open question is whether clean hydrogen can reach a few dollars a kilogram and scale to industrial volumes.

Next: IEA's next annual inventory can test whether the more than 2.5 GW under construction for 2026 actually enters service.
ScalingacceleratingTen-gigawatt installed baseBuild the first 10 GW

Electrochemical Ammonia

Small cells can now make ammonia from nitrogen reliably for days, but they are still far too slow and energy-hungry to replace an ammonia plant.Verified lithium-mediated cells now combine high selectivity with continuous operation, but no system has simultaneously demonstrated industrial current density, energy efficiency, scale and lifetime.

Next: An independently measured integrated stack must pair industrial rate with high energy efficiency, selectivity and long life.
Lab demoacceleratingReach 400 mA/cm² ammonia partial currentMake ammonia about ten times faster

Enhanced geothermal

Hot rock is everywhere underground. The race is to build the plumbing to reach it - cheaply, safely, and at the scale of a real power plant.The physics of engineered reservoirs is proven; the open question is whether they can be drilled cheaply, run for years, and financed at utility scale.

Next: Cape Phase II must show its 21-day representative result across a fleet and disclose full completed-well cost.
EngineeringacceleratingRepeatable fast drillingDrill many wells quickly

Nuclear Fusion

We have made a tiny star flash in a laboratory; now we have to turn that flash into a dependable power station.Fusion has crossed scientific breakeven, but an economical power plant still requires a much harder engineering breakeven.

Next: SPARC targets Q>10, 2027
EngineeringacceleratingEngineering breakevenMake the whole power station pay back its energy

Biology & Health

4

100-Day Vaccines

COVID took 326 days. The target for the next pandemic is 100 - a vaccine before the outbreak spirals.COVID compressed vaccine development from years to months; the goal is a repeatable 100 days from a new pathogen to authorized shots.

Next: CEPI 3.0 must publish viral-family coverage and demonstrate that a stored prototype can accelerate a novel related-virus response.
EngineeringsteadyResponse-Ready Prototype LibraryPrepare before the outbreak

Healthy Longevity

We can slow parts of aging in animals; proving that people gain many healthy years will take long, careful trials that are only just starting.Multiple interventions now reproducibly slow aging in mice, but no treatment has yet shown a large, durable extension of healthy lifespan in humans.

Next: Regulatory qualification of DunedinPACE-class clocks as surrogate endpoints in interventional trials
Lab demosteadyValidated aging biomarkersMeasure aging before decades pass

Speech BCI

Electrodes read the brain's signals for speech. The race is to make them fast, accurate, and durable enough for daily life.Implanted electrodes can already turn neural activity into words on a screen; the open questions are speed, vocabulary, and whether it lasts for years.

Next: Achieve under-10% listener WER on unseen spontaneous sentences with independently controlled pitch and emphasis.
EngineeringacceleratingReal-time expressive voiceRestore a natural voice

Xenotransplant

Thousands die waiting for organs. Gene-edited pig kidneys and hearts could close the gap - if they last.Gene-edited pig organs have kept people alive for months; the open question is whether they can replace human donors for years, not weeks.

Next: Publish a verified 365-day graft-function result, including GFR, rejection episodes and immunosuppression.
EngineeringacceleratingSustain one dialysis-free yearLast a full year

Materials

2
Banked ✓ Live push Later gate Stalled

The whole tree · figure 1 of 2 · the approachfigure 1 of 2

Distance from the centre is the answerThe closer to the middle, the closer to real

Rings are maturity stages; sectors are fields. The bullseye is deployment, and it is honestly empty. Select a problem to light its direct dependencies.Everything heads toward the middle. Nothing has landed yet — tap a problem to see what it needs and what it opens.

5 fields · 20 problems
closer to centre = closer to deployed↓ next — Figure 2 · The Interchange

The whole tree · figure 2 of 2 · the interchangefigure 2 of 2

Five lines, one unbuilt terminusFive lines, all heading for the same station

Each field is a transit line running through maturity zones. Dashed transfers are dependencies; every final stretch into Deployed is unbuilt track.Read it like a subway map. Every colour is a field; the part where it actually arrives has not been built yet.

solid = travelled · dashed = unlaid