Gonçalo Esteves / building capable systems Field Notes →
Space Race 3.0 · Episode 3

The View from Above: satellites and our planet

I read chapter 9 of Red Moon Rising and recorded my questions. Then we researched every one. Here is what we found, what has changed since the book was written, and what we are doing about it.

35questions answered
29book sources checked (notes 41-69)
6still hold as written
23dated, superseded or contested
1dead link

Download the source data (JSON)

What we are doing about it

The book gives a framework. The times are different. For each part of the chapter: what changed, what AI changes, what we are building, and where a small company fits.

Satellites can be listened to, jammed and hacked (notes 41-46)

The book said

The book shows that satellite links were left open (an enthusiast watching NATO spy pictures in 2002) and that Russia attacked the Viasat KA-SAT network on the first day of the Ukraine invasion, while Starlink held up.

What changed

The US and EU publicly blamed Russia for the KA-SAT attack in May 2022, and US agencies issued a warning to satellite network providers. Russia's jamming of Starlink over Ukraine is a continuing, documented contest, tracked each year by the Secure World Foundation.

What AI changes

AI cuts both ways: it helps defenders spot strange traffic on modems and ground stations, and it speeds up the search for bugs in open-source flight software.

What we are doing

We counted which open-source space software projects publish a list of what is inside them (a software bill of materials). One of nine did. We are now building that list for NASA's core flight software ourselves.

Where a small company fits

Small companies can sell software inventories, key management and ground-station hardening to small satellite operators. The US warning gives buyers a ready checklist.

The view from above: climate, El Niño and the sunshade idea (notes 47-55)

The book said

The book uses satellite tracking of El Niño and the long climate record to argue that space gives us the planet's history, then floats a sunshade at the L1 point between Earth and Sun to block a little sunlight.

What changed

NOAA issued an El Niño Advisory on 8 Oct 2026. The sunshade has grown from one-off papers into a small field: a 2025 preliminary design of a precursor mission and the first international sunshade workshop in May 2026. None has flown.

What AI changes

AI makes the satellite archives searchable by people who are not climate scientists. For a sunshade, the hard part is steering a swarm of thin sails, which is a simulation and control problem.

What we are doing

In the episode we ask the question nobody in the papers answers well: if someone can dim the Sun, who decides, and who can switch it off?

Where a small company fits

Earth-observation analytics sold to insurers, farmers and utilities. A strong El Niño year is when buyers look for better forecasts.

Space weather: Carrington, Quebec and the next big storm (notes 56-57)

The book said

The 1859 Carrington storm and the 1989 Quebec blackout show that a solar storm can take down a power grid, and could again.

What changed

How big Carrington was is now given as a range, not one number (USGS, 2024). The May 2024 storm reached G5, the top of NOAA's scale. NOAA's new solar-wind satellite, SOLAR-1 (launched as SWFO-L1 on 24 Sep 2025), reached its L1 orbit in January 2026.

What AI changes

AI models are being trained to predict solar eruptions earlier, but warning time is still set by physics: we see the eruption, then measure the solar wind at L1, roughly 15 to 60 minutes before it reaches Earth.

What we are doing

Our TechLeap entry measures air density in low orbit, which is exactly what a solar storm changes. We tie the two together without claiming more than the instrument can show.

Where a small company fits

Drag and storm-alert data for satellite operators, and outage planning for companies on the ground.

Impacts and planetary defense (notes 58-69)

The book said

Chicxulub, Tunguska and Chelyabinsk show the hazard. Congress ordered a survey in 2005, and surveys like Catalina are the response.

What changed

The 2005 law's goal (find 90% of near-Earth objects 140 m and larger by 2020) was missed. Catalina still leads on spotting small asteroids hours before they hit: 2026 RW1 was seen about six hours before it fell into the Indian Ocean in September 2026. A 2024 study traced the dinosaur-killer to a carbon-rich asteroid from beyond Jupiter.

What AI changes

Software is the bottleneck, not telescopes: a tracking algorithm run on old archive images found about 27,500 new asteroid candidates in 2024. That was clever search code, not AI.

What we are doing

We time an episode to ESA's Hera arriving at the asteroid DART hit. The number we want to build: how many impacts were caught before they happened, year by year.

Where a small company fits

Orbit software and follow-up observing on public telescope archives. The missed 90% goal is a documented, unmet mandate.

My questions, and what we found

Every claim below links to a source we opened on 3 Nov 2026. Company figures are the company's own claims.

A · Earth weather from space: the data and who pays

A1What does NOAA do, does it publish its data in a repository, and can the data be mined for predictions?

Yes, but not mainly on GitHub. NOAA puts its data on Amazon, Google and Microsoft clouds through its Open Data Dissemination program. That includes forecast models such as HRRR and satellite data from GOES-16 to GOES-19. GitHub holds the code: the NOAA-EMC account hosts the workflow behind its global forecast systems (GFS, GEFS, SFS) under an open licence. Anyone can build on both. NOAA's own forecast is the free starting point.

What you could build: Sell a local or sector layer on top of the free model output: alerts tuned to one business decision, not another copy of the national forecast.

A2Which weather and Earth-observation data sources are open, which are paid, and how do you connect them?

Open: NOAA's cloud datasets; ECMWF's real-time IFS and AIFS forecasts, free under a CC-BY-4.0 licence that allows commercial use with credit; NASA Earthdata, under a full and open sharing policy; and Copernicus Sentinel data, free through the Data Space Ecosystem. Paid: Planet sells daily soil water content and land surface temperature at 1 km, and crop biomass at 10 m. Hydrosat sells thermal infrared data and analytics. Start free, then buy the one paid feed that fills a gap.

What you could build: A connector that pulls the free feeds into one customer dashboard, then adds a single paid layer the customer actually needs.

A3My phone gives me a good-enough forecast in seconds. Who pays for something better, how much, and for what?

Buyers pay for data tied to a decision. NOAA buys radio-occultation data: in March 2023 it signed two five-year contracts with PlanetiQ and Spire worth up to $59.3 million in total. Planet says it supplies soil water content data in 17 countries, working with reinsurers to protect farmers from drought. Tomorrow.io's Series F reached $210 million, with an insurer, Harel, in the round. Hydrosat raised €51 million ($60 million) in January 2026 for field-level crop data.

What you could build: Pick one paying decision (irrigation, crop insurance, outdoor work) and sell the forecast that answers it, with its error shown.

A4Which company gives farmers precise soil moisture from space (the host recalled "Space Tangle"), and how precise is it?

The name is a mix-up. Space Tango is a Kentucky microgravity research and manufacturing company. It has supported over 300 experiments on about 40 missions to the Space Station. On 25 June 2026 Kentucky's governor announced a $7.4 million expansion of its Lexington headquarters. Soil moisture from space is sold by others. Hydrosat sells field-level soil moisture and crop data from thermal satellites. Planet sells soil water content near-daily at 1 km resolution, built on passive microwave data.

What you could build: Ground-truth service: check satellite soil moisture against in-field probes for growers and insurers, and publish the error.

A5What severe phenomena does space now track, El Niño and what else, and who pays for the tracking?

On 8 October 2026 NOAA's Climate Prediction Center kept its El Niño Advisory, with Niño-3.4 at +2.1°C. It gave a better than 83% chance of a strong-to-very-strong event through January to March 2027. Wildfire is the newest lane: the first three operational FireSat satellites launched on 7 July 2026. FireSat is backed by the Bezos Earth Fund, Google.org and the Moore Foundation. NOAA also tracks space weather, as in the January 2026 radiation storm.

What you could build: Turn public El Niño and fire alerts into plain, local briefings for farms, utilities or insurers.

A6Does GPS radio occultation feed long-range climate models, who sells it, and can it model 300–500 km for drag?

Radio occultation is real and useful. UCAR lists three uses: weather forecasting, climate monitoring and model checks, and space weather. Its profiles give temperature, pressure and water vapour in the air, and electron density in the ionosphere. About 80% reach weather centres within 3 hours. We found no source for the book's claim that it supplies most climate-model data. Spire and PlanetiQ sell it to NOAA. Air density at 300–500 km, the drag number, is not among those products.

What you could build: A drag product needs its own density data; radio occultation can be a cross-check, not the source.

B · Space weather

B1Can we chart the major solar storms, and where do the book's storm probabilities come from?

Key storms: 1859 Carrington; 1989, when Québec lost power for nine hours; October 2003; a near miss on 23 July 2012 (a CME at 2,500 km/s, aimed away from Earth); and January 2026, when a radiation storm beat 2003. A 2024 USGS study puts Carrington at about -866 to -964 nT. The odds are model outputs: a 2017 study gave 10.3% per decade under one statistical fit, 3.0% under another. Lloyd's cited a Carrington return period of about 150 years.

What you could build: A plain, sourced storm-odds explainer for operators, showing the range rather than one number.

B2Carrington's CME hit about 18 hours later. What physics sets arrival time?

Arrival time is distance divided by average speed. Carrington's CME is estimated to have covered about 150 million km in 17.6 hours, an average near 2,400 km/s. NOAA says CMEs range from under 250 km/s to nearly 3,000 km/s. The fastest Earth-directed ones arrive in 15–18 hours; slower ones take several days. One explanation for Carrington's speed: an earlier CME may have cleared the way.

What you could build: An arrival-time calculator that shows the uncertainty window for one customer's assets.

B3Beyond the Sun, what is space weather: cosmic rays and more? Effects at GEO, LEO and the Moon, and what we can do.

Three things hit hardware. Low-energy electrons charge spacecraft surfaces. High-energy electrons charge the inside. Solar particles and galactic cosmic rays, from outside the solar system, cause single-event upsets in electronics and add up total dose (NASA). Geostationary satellites face extra risk in big radiation storms, NOAA said in January 2026. For crews near the Moon, NASA's Orion carries six radiation sensors, and a high dose reading triggers a shelter recommendation.

What you could build: Radiation and charging checks for small-satellite builders, written for owners rather than physicists.

B4Are DSCOVR, ACE and SOHO still working in 2025–26, what replaced them, and is ESA's Vigil coming?

The fleet changed this year. NOAA's SOLAR-1 (launched as SWFO-L1 on 24 September 2025) reached L1 on 23 January 2026 and went operational on 10 June 2026. NOAA's space weather centre now uses SOLAR-1 as its primary solar-wind source, with ACE as backup. It has stopped processing DSCOVR data. So DSCOVR is no longer NOAA's main L1 source. ESA lists SOHO, launched 1995, as operational. ESA plans to launch Vigil to L5 in 2031.

What you could build: Downstream tools that read the new SOLAR-1 feed and translate it for satellite and grid operators.

B5How much warning do we get, and who holds the information?

Fast CMEs reach Earth in 15–18 hours; slower ones take several days. The last, firmest warning comes from spacecraft at L1. NOAA says they often give 15 to 60 minutes' notice before the shock arrives. NOAA's Space Weather Prediction Center is the official US source of space weather alerts and warnings, as CISA notes. ESA says Vigil, planned for 2031, could give up to four to five days' notice.

What you could build: An alert service that turns the public warning into a 'what to do in the next hour' checklist for one asset type.

B6What is CISA doing on space weather and space assets, and who owns protection?

CISA's risk centre co-chairs SWORM, the federal space weather coordination group. It leads updated storm benchmarks and works with governments, insurers and the private sector. CISA's page says space weather, unlike an EMP attack, is inevitable. The PROSWIFT Act (2020) set up a federal space weather advisory group. The first national end-to-end space weather tabletop ran on 8–9 May 2024. Ready.gov notes storms can induce currents that disrupt power lines.

What you could build: Help small infrastructure owners map their exposure to the federal benchmarks.

B7Can I insure my house or car against a solar storm, and how do companies with ground assets insure?

Untested for homes. Progressive says standard policies may cover power surges from lightning, but often not surges from faulty wiring. The page does not mention geomagnetic storms. Lloyd's 2013 report said a Carrington-level storm could cut power to 20–40 million Americans for weeks to two years, costing $0.6–2.6 trillion. Lloyd's 2025 scenario puts global losses at $2.4 trillion over five years, with an expected loss of $17 billion. We found no source on car or military cover.

What you could build: A policy-wording check that shows a business owner which storm losses their cover would and would not pay.

B8What if a company loses phones for hours, or malware takes systems down? Which scenarios should it plan for?

Start with the free tools. CISA's Tabletop Exercise Packages are customisable drills with objectives, scenarios and discussion questions, covering threats from natural disasters to industrial control systems. Ready.gov's business pages cover communications, IT recovery and continuity planning, plus home outage preparation. Big operators already drill: over 370 organisations joined the GridEx VIII grid exercise in November 2025. The EIS Council runs Black Sky drills for long, widespread outages. CISA notes space weather is a different threat from EMP.

What you could build: Half-day outage drills for small firms that are not in GridEx, built on CISA's free exercise packages.

B9What must Artemis astronauts watch for in space weather compared with Apollo?

For Artemis II, a NASA Goddard office tracks every solar eruption and passes its analysis to the radiation team at Johnson. NOAA adds forecasts. NASA even used the Perseverance rover on Mars to spot large sunspots up to two weeks before they face Earth. Inside Orion, six sensors and personal dosimeters track dose. One threshold triggers closer watch; a higher one triggers a shelter recommendation. We found no source comparing this with Apollo, so we leave that open.

What you could build: A forecast and dose-alarm service for commercial lunar landers and payloads.

C · Lagrange points and the sunshade

C1Why do spacecraft park at Lagrange points, how many are there, and do they need station-keeping?

NASA counts five Lagrange points, where a small object keeps pace with two large bodies. L4 and L5 are stable. L1, L2 and L3 are not: ESA likens them to a ball on a hilltop, so spacecraft fly halo orbits and fire thrusters often. Webb is not parked: it circles L2 every 168 days, between about 250,000 and 830,000 km from the point. NASA said in 2021 it had fuel for well over ten years.

C2Who owns the parking rights at a Lagrange point, and who decides what you can do there?

No one assigns slots. The Outer Space Treaty says outer space is not subject to national appropriation by claim of sovereignty, use or occupation (Article II). It also asks states to act with cooperation and due regard (Article IX). The ITU recommends a 250,000 km radio-quiet coordination zone around L2, but that covers radio, not position. A 2025 law journal article argues Lagrange points should be treated as the common heritage of mankind.

What you could build: Regulatory and frequency-filing help for small teams planning missions to L1 or L2.

C3How big is a Lagrange point region, could it hold a station, and does it move?

The point is a spot in the maths; spacecraft use the region around it. Webb's loop around L2 ranges from about 250,000 to 830,000 km from the point. The ITU's protected zone around L2 has a 250,000 km radius. The points move with Earth: a spacecraft at L2 keeps pace with Earth's orbit around the Sun (ESA). Any station at L1 or L2 would need the same frequent thruster firings as today's spacecraft. We found no planned station there.

C4A sunshade at L1 to filter sunlight: who proposed it, what would it cost, and who would control the weather?

The best-known design is a 2006 PNAS paper from the University of Arizona. It would block 1.8% of sunlight with a cloud of one-gram 'flyers', about 20 million tons in total. They would launch in stacks of 800,000 and fill a cloud 100,000 km long. It estimated about 25 years and a few trillion dollars. On control: no body governs it. The Outer Space Treaty bars any nation from claiming outer space, but sets no rules for shading Earth.

D · Planetary defense

D1Is Spaceguard still working? The 1 km goal was met in 2011. What is the 140 m target now?

The 1 km goal was met: in 2011 NASA's WISE data showed at least 90% of large near-Earth asteroids found. The 2005 law then set 90% of objects 140 m or larger within 15 years, so by 2020. That was missed. A 2023 study put completeness at about 38% for 140 m and up. NASA's database lists 43,047 known near-Earth objects today. NEO Surveyor, built to close the gap, launches no earlier than September 2027.

What you could build: Follow-up observation and orbit software that sells into the gap Congress set but did not fund on time.

D2Who looks for dangerous asteroids: PDCO, Catalina, ATLAS, Pan-STARRS?

NASA's Planetary Defense Coordination Office funds the surveys. Catalina Sky Survey (University of Arizona) is NASA-funded and has found 17,940 near-Earth asteroids since 1995, 663 of them in 2026. ATLAS (University of Hawaii, NASA-funded) runs telescopes in Hawaii, Chile and South Africa, and reported data from a Tenerife unit in June 2026. NASA describes Pan-STARRS1 as funded by its NEO Observations program. The Rubin Observatory added over 11,000 new asteroids in April 2026.

What you could build: Software and data services for survey teams; the telescopes are grant-funded, the tools around them are open to small firms.

D3What if 'Oumuamua was a spacecraft? What are we tracking now?

Pan-STARRS1 found 'Oumuamua on 19 October 2017. In 2019 a team of 14 experts concluded it has a natural origin. In 2023 NSF-backed researchers explained its odd push as hydrogen escaping from ice. A 2025 preprint argued 3I/ATLAS could be technology, but called itself largely a pedagogical exercise. NASA calls 3I/ATLAS a comet; it posed no threat and passed closest to the Sun around 30 October 2025. 2I/Borisov, found by an amateur in 2019, was the first confirmed interstellar comet.

D4Does B612 still run open-source tools, and which open asteroid data repositories exist?

Yes. B612's Asteroid Institute runs ADAM, with tools for precovery (finding objects in old images), impact probability, trajectory design and ephemerides. It publishes open code (ADAM Core, Precovery, THOR, MPCQ, Quivr) and a free BigQuery copy of Minor Planet Center data. NASA JPL offers free APIs: one returns lists and counts of small bodies, another the fireball record. These are the main open doors for an outside team.

What you could build: Run the open orbit-linking code on new survey data and offer candidate vetting as a service.

D5Can AI find objects we have missed in the archives, and what is the Earth-like-planet project mentioned on Moonshots?

Partly, and the famous number was not AI. The 27,500 new asteroid candidates (April 2024) came from THOR, a tracking algorithm, run over 5.4 billion archived observations. They are candidates, including over 100 near-Earth ones. B612 said it was exploring AI only to vet candidates. For planets, AI is real: the deep-learning model ExoMiner++ sorted 147,568 TESS signals and flagged 7,330 planet candidates. We found no XPRIZE page for a habitable-worlds prize.

What you could build: Better search over public archives: candidate vetting, ranking and follow-up lists.

D6Do we really know what is out there, what is withheld for national security, and do operators want the data?

Some is public. NASA's fireball database comes from US government sensors; CNEOS does not verify these events. Today it holds 1,074 events, from April 1988 to October 2026. In 2022 US Space Command confirmed a 2014 meteor came from outside the solar system. Researchers had trouble publishing that finding because it relied on classified data. We found no published demand signal from commercial operators.

What you could build: Cleaned, cross-checked fireball catalogues and alerts; demand from operators is still to be tested.

D7Are we insured against an asteroid impact, the way we are for earthquakes or tornadoes?

Probably yes by accident, not by design. After the 2013 Chelyabinsk meteor, catastrophe modeller AIR Worldwide said broad multi-peril policies generally cover all risks not specifically excluded. So meteorite damage would generally be covered, and so would falling objects if not excluded. We found no asteroid-specific insurance product or catastrophe model. NASA's 2024 YR4 scare ended with NASA saying the object poses no significant impact risk to Earth in 2032 and beyond.

What you could build: An airburst loss model built from NASA's open impact data; no insurer has validated one yet.

D8DART proved we can hit an asteroid. What happened to the debris, and why isn't deflection predicted better?

DART hit the 160 m moonlet Dimorphos on 26 September 2022 and cut its orbit by 33.0 ± 1.0 minutes. The debris mattered most. Dimorphos got 2.2 to 4.9 times the momentum DART carried (the factor called beta), because ejecta added to the push. The range is wide because Dimorphos's mass was unknown. That is why prediction is hard. ESA's Hera is due at the asteroids in November 2026 to survey the result.

What you could build: Simulation and target-characterisation tools: mass, shape and strength from cheap fly-bys.

E · NASA science

E1What is NASA science funding now, what is the Administrator pushing, and what changes for planetary defense and space weather?

FY2026 Science was $7.25 billion. The FY2027 request, released 3 April 2026, would cut Science 46% to $3.89 billion and NASA 23%. The Planetary Society counts 53 missions likely ended, 17 of them in heliophysics, the science of the Sun and its effects. Stopgap funding runs flat to 11 December. On 21 October NASA's science chief wrote that the Administrator directed faster science through a new Competition Model 2.0. NEO Surveyor remains scheduled for no earlier than September 2027.

What you could build: Shorter selection cycles favour small, first-time teams; plan on flat funding until the December deadline passes.

E2Who is studying how Mars lost its atmosphere?

The host's framing needs an update: MAVEN, the orbiter that measured the loss, is gone. NASA lost contact on 6 December 2025 and declared the mission over on 3 June 2026, after more than 11 years at Mars. The active mission is ESCAPADE: two identical spacecraft launched 13 November 2025. They wait in a loop near the Sun-Earth L2 point. Then they slingshot past Earth in fall 2026 to study how the solar wind strips Mars's atmosphere.

What you could build: ESCAPADE shows the low-cost model: two small spacecraft and a commercial launch.

F · Earth weather: can we control it?

F1Do we have enough prediction models, who pays for accurate weather, and what does accuracy mean?

Models are not the bottleneck. NOAA launched three AI models on 17 December 2025. One uses up to 99.7% less computing. Another adds 18 to 24 hours of forecast skill over the traditional ensemble. ECMWF's forecasts are free for commercial use with credit. Accuracy means a score against what actually happened, such as track error or extra lead time. Governments pay for the base. NOAA also buys commercial radio-occultation data under contracts worth up to $59.3 million.

What you could build: Independent verification: score free forecasts against local outcomes for one customer, and sell the trust.

F2How does cloud seeding work, who does it, and is there a drone fleet that can make rain within hours?

Seeding releases silver iodide into clouds. Ice crystals form on the particles and grow until they fall as rain or snow. Rainmaker (El Segundo) flies drones into winter storm clouds and measures results with radar. The company says it logged 82 seeding signatures in four months, and 19 million gallons near Homer, Alaska, in about three hours. NCAR's 2017 Idaho study found seeding created snowfall. We found no source for a company promising rain in 5–7 hours.

What you could build: Independent measurement of seeding results for water districts.

F3What if cloud seeding goes wrong, a superstorm? What are the effects on climate and biodiversity?

On the evidence found, seeding effects are modest. After the July 2025 Texas flood, Rainmaker said its best operations produce tens of millions of gallons, at most about 100 million. The flood dropped over a trillion gallons. A climate scientist in the same report said that in some cases it is hard to prove seeding does anything. We found no study on downwind or biodiversity effects. That gap is the real risk.

F4Who owns the rain? Is it legal to take rain meant for somewhere else, and can places trade quotas?

Federal law only requires reporting: anyone seeding clouds must report it to the Commerce Department (15 CFR 908, from a 1971 law). The rest is state law, and it varies. Florida repealed its licensing law and banned certain acts meant to change weather or sunlight, effective 1 July 2025. Alaska's natural resources department said cloud seeding is not addressed in state law. In October 2026 the Kenai Peninsula Borough created a permit rule. We found no law setting rain quotas.

What you could build: Permitting and reporting help for seeding operators moving state to state.

F5Is weather-control tech ready? The AI angle, the parts list, the business case.

Partly. Rainmaker flies drones, releases silver iodide, and uses radar and validation systems to estimate the extra precipitation. Since July 2025 it has partnered with Atmo, an AI weather forecasting startup, to pick clouds worth seeding. Rainmaker expects its water cost for November 2026 to April 2027 to fall below desalination; it gave no figure. It flies into winter storm clouds, so those clouds must already exist. We found no public parts list or unit costs.

What you could build: Flight operations, validation and permitting services around seeding, rather than the drones themselves.

G · Constellations and software bills of materials

G1Constellations and SBOMs: who are the satellite operators, what software do they run, and is there a business?

Starlink dominates: 11,148 of 17,099 active satellites on 26 October 2026, about 65%. Operational counts for others: OneWeb 625, Amazon Kuiper 368, Xingwang 224, Qianfan 195. A SpaceX lead said in 2020 that Starlink runs Linux and only software SpaceX has signed. So SpaceX is a reference point, not a buyer. In July 2026 CISA, NSA, FBI and international partners published new minimum SBOM elements, adding an author signature.

What you could build: SBOM checks for small operators and suppliers whose parts fly on big constellations, timed to launches, contracts and insurance renewals.

The chapter's sources, checked

Endnotes 41-69 of chapter 9: is the source still there, what kind of source is it, does it still hold, and what is newer. Our one-line notes, not the book's words. Download the data (JSON).

#SourceLinkTypeStill holds?Newer
41 Urban, BBC Newsnight 2002: enthusiast watches NATO spy pictures online secondary dated
HTTP 200 (BBC archive page). 2002 anecdote of an amateur receiving unencrypted satellite feeds. Still a valid illustration of open links; modern versions are documented in Pavur's 2020 work.
42 Wess, USNI Proceedings Feb 2021, ASAT Goes Cyber online secondary dated
HTTP 403 (bot block, not read). Professional-journal opinion piece by a Navy officer; USNI author page confirms the byline. Pre-dates KA-SAT, so 'cyber ASAT' is now demonstrated rather than predicted.
43 Space Attacks Open Database, spacesecurity.info, accessed 2023 online secondary dated
HTTP 200. Page shows its latest entry as 2019 (Microsat-R test), refs accessed July 2019: the book accessed it in Nov 2023 but the database looks unmaintained; it omits KA-SAT and 2021-26 events.
44 Blinken, State Dept 10 May 2022: attribution of Russian cyber activity moved primary yes
URL 404 as written; page exists at the same path without the trailing slash (state.gov, not fetched: 403 bot block). Coverage confirms the 10 May 2022 US statement with UK and EU on the KA-SAT attack. Note: the US statement blames Russia; SentinelOne found only loose code ties to a Russian actor and no formal attribution.
45 Musk tweet, 10 May 2022, Starlink resisting Russian jamming and hacking online social dated
HTTP 200 in the automated check, not re-read. Reported wording: Starlink 'has resisted Russian cyberwar jamming & hacking attempts so far, but they're ramping up their efforts'. A company owner's statement, not independent proof.
46 Horton, Washington Post 18 Apr 2023: Russia tests weapon against Starlink online secondary dated
HTTP 000/403 (paywall or bot block, not read; byline unconfirmed). Based on leaked US documents (Tobol-1, 'ongoing military operational experiment'); outages have not been tied to Tobol vs other jammers. Ukraine reported Starlink disruptions from May 2024; Russian-media claims about 'Kalinka' are secondhand.
47 Becker, NOAA ENSO Blog, 5 Mar 2015: El Niño is here online primary dated
Valid for 2015 event only. Book uses it as an example of satellite-era ENSO tracking. NOAA CPC issued an El Niño Advisory on 8 Oct 2026 (Niño-3.4 +2.1 C; >83% chance strong-to-very strong through JFM 2027). Book's URL: author blog, the 'Mach 5' in the citation is a typo for March 5.
48 Autry, Foreign Policy, 20 Jul 2019: space research and climate online author-own dated
Opinion piece by the book's own author; argues Earth-science from space. Claims are his framing, not independent evidence. 2019; no climate-satellite numbers should be taken from it today.
49 Encyclopedia Britannica, 'Paleoclimate' (Cretaceous), accessed 15 Nov 2023 online secondary yes
Tertiary reference; 403 is a bot block. Fine as background; replace with a primary paper (e.g. the Chicxulub or KPg work) if used on air.
50 Jensen et al., Boreas 1999: Ancylus Lake stage, SW Baltic print academic dated
Journal article, no URL; 1999. The Baltic freshwater-lake history is standard, but we did not open the paper. Check for a DOI via Boreas before citing.
51 Herrle et al., Sci. Reports 2018: Black Sea outflow, Holocene meltwater print academic contested
Sci. Reports 8:4081 (2018), not opened today. Supports gradual-outflow readings; the Black Sea 'flood' itself remains contested (Aksu 2002, 2022 review favour gradual outflow; Lericolais 2007 argues for flood).
52 Ballard et al., Am. J. Archaeology 2001: Sinop deepwater archaeology print academic contested
2001 survey found shoreline/wreck evidence; a 2011 review found no archaeological evidence for catastrophic submergence of settlements (seen via search, not opened). Do not say the flood is proven.
53 Cornish, NASA: What is a Lagrange point? moved primary yes
Physics unchanged. Old solarsystem.nasa.gov URL returns 301 to https://science.nasa.gov/resource/what-is-a-lagrange-point/ (observed 3 Nov 2026). Update the footnote.
54 Angel PNAS 2006 and Fuglesang 2021 Acta Astronautica: L1 sunshades print academic dated
Both real papers (not opened today). Angel's concept is 1.6 million flyers; Fuglesang's targets ~1% sunlight cut. Both are concepts, none flown. Newer: 2025 Acta Astronautica precursor-mission design; first IAA Planetary Sunshade Workshop held 13-15 May 2026 (Nottingham); a 2026 roadmap paper. Fuglesang does not appear as author on the 2025/2026 items we saw.
55 Bromley, Khan, Kenyon, PLOS Climate 2023 (lunar dust shield); Forbes 2023 online academic contested
Peer-reviewed model; Forbes 403 is a bot block. Authors themselves say it is no substitute for cutting emissions; amount of dust is mining-scale; critics call it speculation. we found no 2025-26 follow-up by them.
56 Green & Boardsen, Adv. Space Res. 2006: Great Auroral Storm 1859 print academic dated
Real paper (not opened). Carrington size is now less certain: Love, Hayakawa et al. 2024 give a median peak Dst of about -866 nT (excl. one datum) or -964 nT, with large uncertainty; a 2026 deep-learning study gives an upper limit near -1313 nT. 'One in a million' tail wording is theirs, not ours. May 2024 Gannon storm (G5, Kp 9, Dst about -412 nT) is the modern reference; SWFO-L1 launched 24 Sep 2025.
57 Phillips, SpaceWeatherArchive, 12 Mar 2021: Great Quebec Blackout online blog dated
Fine narrative of 13 Mar 1989; 2021 blog, not primary. Use Hydro-Quebec/NERC post-mortems or SWPC for numbers. Spaceweather.com site is active (mirror showed May 2026 entries). Space weather is now in the news cycle: Solar Cycle 25 maximum declared 15 Oct 2024 by NOAA/SWPC panel.
58 Schulte et al., Science 2010: Chicxulub and the K-Pg extinction print academic dated
Landmark review (Science 327:1214), not opened. Newer: Aug 2024 Science paper (ruthenium) says the impactor was a carbonaceous asteroid from beyond Jupiter, not a comet.
59 Amos, BBC, 18 Nov 2016: 'Instant Himalayas' (Chicxulub peak ring) online secondary dated
Reports IODP/ICDP drilling result. Still valid as a pointer; 2016.
60 Amos, BBC, 15 May 2017: asteroid hit 'worst possible place' online secondary dated
Reports the Chicxulub hydrocarbon-rich target rock finding (Kring et al.). Still the standard line.
61 Robertson et al., GSA Bulletin 2004: first hours of the Cenozoic print academic contested
Argues global ejecta-heated 'broil'; the thermal-pulse scale is debated in later work. Not opened today.
62 Napier & Clube, Nature 1979: A theory of terrestrial catastrophism online academic dated
DOI resolves (200). Early cometary-catastrophism case, 1979. Historical, not current science.
63 Space.com, 'Nemesis Star Theory' (accessed 5 Jan 2024) online secondary superseded
Hypothesis of a dwarf-star companion: NASA WISE (Luhman 2014) found no Saturn-mass or larger object to 10,000 au, no object larger than Jupiter to 26,000 au. A faint substellar variant is not fully excluded (reviewer's wording). Say 'not found, unlikely', not 'disproved'. Space.com page body would not load for us.
64 Popova et al., Science 2013: Chelyabinsk airburst print academic yes
Still the reference paper for the 15 Feb 2013 airburst (not opened today). A June 2025 article reported origin in the Flora family (secondary, date unclear).
65 RBC (Russia), 18 Feb 2013: Chelyabinsk injured near 1,500 online secondary yes
401 from my fetch: block/auth, not dead. Number (about 1,500 injured, mostly from glass) matches the widely reported figure. News source, Russian-language; cite alongside Popova 2013 or ESA/NASA.
66 Farinella et al., A&A 2001: Tunguska body's asteroidal origin print academic contested
2001 dynamical argument. A 2020 preprint (Khrennikov et al.) proposes an iron grazer that left the atmosphere; Mark Boslough doubts its shock-wave physics. Unsettled.
67 Pew, Jul 2023: Americans' views of space, NASA priorities online primary dated
Survey of 10,329 adults, 30 May-4 Jun 2023. Top NASA priority: monitoring asteroids; only 12% said crewed Moon landing top priority; 47% said private companies make important contributions, 40% unsure. Still Pew's newest dedicated space survey as far as we found. Opportunity: public ranks planetary defense first, budget does not.
68 Chapman 1998 House testimony; NASA Authorization Act 2005 (George E. Brown NEO Survey) dead link primary dated
Chapman URL returns 404 on boulder.swri.edu and www.boulder.swri.edu (3 Nov 2026); archive.org blocked for us, so no archive URL. govinfo PLAW-109publ155 loads (200). The Act's Sec. 321 survey goal (from the Act text, which we did not re-read today) was 90% of NEOs 140 m+ by 2020; that goal was not met, and a 2025 NASA OIG report says NEO Surveyor slipped from 2026 to 2028 and from $1B to $1.6B (via search summary, verify). NEO Surveyor now NET Sep 2027 per one May 2026 aggregator (NASA page not found).
69 Catalina Sky Survey, LPL Univ. of Arizona (accessed Jul 2023) online primary yes
Site loads. Survey is still the lead for pre-impact detections: 2026 RW1 (6 Sep 2026, about 6 h warning, Indian Ocean; reported as the 13th detected before impact, secondary count).

How we work

We report what has been reported, with who said it and what kind of source it is ("this is what we know"). Government and company claims are labelled as claims. We never publish how to attack a system. Spot an error? Tell me and the correction goes on this page with its date.

Other episodes

EP00 · The map · EP01 · What is inside open-source space software

Built from the research files by script, checked 2026-11-03. — Gonçalo Esteves