Tom Elwood -best of the internet.
Here are some youtube videos, or articles that caught my eye - from the New York Times, Consumer Reports, Popular Science etc.
Monday, November 16, 2026
Why Iran can't build a nuke
The intrigue of an interpreter who was simply given far too much access to secure documents about building nuclear warheads, and used it to sell revenge on his country's enemy.
15:45 In order "to get enough U-235 to build just one warhead, you need about 1000 centrifuges running 24/7 for about a year"
The world's biggest EV push
Ethiopia has banned sales of all but electric vehicles in a sweeping and rapid change. They have an ambitious goal of infrastructure build-out of charging stations to keep up with the resulting demand.
This initiative arose at a time when they were crippled in debt trying to pay $5B a year for oil, when they instead spent $5B on building a hydroelectric dam of the Nile River to allow 96% of their energy to be generated cleanly, without fossil fuels.
But EV's are very expensive, over 25x the average annual income.
12:50 car ownership is 1.3% inhabitants in Ethiopia vs 7.3% in the rest of Africa.
This transition to EV's will be hard to reproduce in other countries; Ethiopia has the perfect set-up of 15:18 a huge hydroelectric resource in the Nile river, a relatively small car market for the population, and a severe fuel crisis indebtedness.
The truth about how data centers use water
When a company claims they use a closed system that doesn't consume water, they're veiling the truth because that must be paired with a system that does use water to dispel the heat. Although it can sometimes be cooled with "dry cooling" using air when the ambient temperature is low, in summer heat, evaporative cooling is needed to efficiently liberate the excess heat.
Future innovations include 11:05 getting cooling water closer to the chip, and 9:51 designing chips that generate less heat from the same computing power.
How do cyber security staff distinguish between a hostile attack vs regular outage?
In light of all the recent outages (Outlook, Air Traffic Control, Chase Bank,) and repeated official claims that they are not hostile cyber-attacks, I wondered how the experts so categorically tell the difference between a normal outage and a cyber-attack. It turns out they look for outside influences, unusual patterns (timing of a software change, source country of requests,) unusual code (e.g. sending pings to foreign computers) and so on.
I've edited this AI response below for brevity.
Cybersecurity investigators, incident response teams, and federal agencies treat every major infrastructure failure as suspicious until proven otherwise. Distinguishing between a mundane software bug and a covert cyberattack comes down to forensic evidence across four main areas:
1. Causation and "Patient Zero"
In a routine technical outage, engineers can trace the failure backward to a specific, authorized change—such as a scheduled firmware update, a misconfigured routing table, or a failing piece of hardware.
Routine Glitch: The crash correlates directly with a known internal maintenance ticket or a sudden spike in normal user traffic.
Cyberattack: Systems fail without an authorized change, or changes are executed using compromised administrator credentials outside normal maintenance time windows—often originating from an unexpected IP address or unfamiliar device.
2. Traffic Patterns and Telemetry
Network logs capture every packet entering and leaving a system, leaving distinct fingerprints:
Distributed Denial-of-Service (DDoS): An attack floods servers with millions of junk requests from thousands of botnet devices globally, looking completely different from legitimate customer traffic.
Data Exfiltration & Beaconing: Covert attackers rarely just break things; they usually steal data or establish persistent access first. Investigators look for outbound data transfers to unknown foreign servers or periodic "heartbeats" (beaconing) where infected machines check in with an attacker's command-and-control server.
3. Indicators of Compromise and "tactics, techniques, and procedures."
Threat hunters scan memory dumps, system registries, and server logs for known malicious artifacts:
Unrecognized Scripts: Unauthorized executables, obfuscated PowerShell commands, or altered system files that shouldn't exist on production servers.
Log Tampering: A major red flag for hostile activity is missing or wiped event logs. Software bugs generate chaotic error logs; attackers intentionally delete logs to cover their tracks.
Tactics, Techniques, and Procedures (TTPs): Agencies match suspicious behaviors against known nation-state or ransomware playbooks (cataloged in frameworks like MITRE ATT&CK).
4. Blast Radius and Targeting Logic
How a failure spreads across a network reveals whether it is accidental or intentional:
Accidental Cascade: A bug or hardware failure spreads indiscriminately along physical or logical dependencies (e.g., a failed DNS server taking down everything connected to it).
Surgical Sabotage: Attackers intentionally target high-value choke points—such as backup servers, Active Directory domain controllers, or industrial control valves—simultaneously across isolated network segments that shouldn't fail together.
Why Attribution Takes Time—or Happens Quickly
When a company like Microsoft or Chase rules out a cyberattack within hours, it is usually because their engineers immediately identified the exact internal code commit or hardware fault that broke the system and watched services recover the moment that change was rolled back. Conversely, if systems refuse to stay up after a rollback—or if logs show unauthorized privilege escalation—agencies like Cybersecurity and Infrastructure Security Agency (CISA) are brought in for deeper digital forensics.
Sunday, November 15, 2026
In-depth analysis of 2 recent events released just 24 hours later
You may have already seen enough about these 2 recent events, or these videos may not be what you're interested in... they're quite long and go into unnecessary detail. But what impresses me is that both videos with extensive animations, re-creations, and analysis came out within 24 hours of the events. The ability of these video producers, assumably with the aid of AI video production, to produce this quality of video in such a short time is remarkable.
But the first one is about the crash and sinking on the Mississippi River last week, with some good observations by a shipping expert on what may have been happening, noting that one anchor failed to deploy properly suggesting a ship-power malfunction, how lucky they were to avoid striking a nearby bridge, and how quickly other ships and tugboats responded to help.
Nationwide ground stop for all aircraft on Nov 13 '26
In this second video, they explain in detail what happened to the air traffic control system, a huge separate dedicated communication infrastructure across the continent, that is being updated from copper wire to fiberoptic. Older devices that process multiplexing of multiple signals got overloaded and broke up the signals of aircraft position and pilot communication so that central air traffic control couldn't rely on the actual positions of aircraft in real time, so they grounded all flights for 4 hours. How they bypassed the overloaded equipment, tested it, and got it back up and running in 4 hours is amazing to me, and they don't go into detail on that. But the level of complexity of the system, its redundancy, and how well it's illustrated in their animations, is remarkable. (I guess it's possible they were already working on a video about the ailing system when this event happened.)
Friday, November 13, 2026
Android sluggishness resolved
A recent Google system update introduced a memory management (ZRAM / RAM allocation) bug on the Pixel 8 that causes severe frame drops, UI stutter, and sudden sluggishness—even when storage is empty and no malware is present.
Google is rolling out a patch to fix this underlying ZRAM issue, but you can immediately restore smooth performance on your Pixel 8 using the steps below.
1. Apply the Immediate Fix (Background Process Limit)
Setting a temporary cap on background tasks bypasses the ZRAM memory-clogging bug:
1. Go to Settings -> About phone.
2. Scroll to the bottom and tap Build number 7 times until you see the message "You are now a developer."
3. Go back to Settings -> System -> Developer options.
4. Scroll down to the Apps section and tap Background process limit.
5. Change it from Standard limit to "At most 4 processes."
(*Note: Developer options occasionally revert this setting after a reboot. If the lag returns after restarting, re-check this limit.)
2. Clear System UI & Pixel Launcher Cache
A corrupted launcher cache can cause the Pixel's swipe gestures and home screen animations to lock up:
1. Go to Settings -> Apps -> See all apps.
2. Tap the three dots in the top-right corner and select "Show system." (This battle increases the number of apps to scroll through.)
3. Search for and select Pixel Launcher.
4. Tap Storage & cache -> Clear cache.
5. Repeat these steps to search under system apps for System UI, Tap Storage & cache -> Clear cache.
3. Toggle "Smooth Display"
The Tensor G3 chip manages variable refresh rates (60Hz to 120Hz). A display driver bug can get stuck rapidly switching frequencies, appearing as lag:
1. Go to Settings -> Display.
2. Toggle Smooth Display Off, wait 5 seconds, and turn it back On.
3. If the lag vanishes, leave it on, or keep it off temporarily if the stuttering persists until the next update.
Thursday, November 12, 2026
Very clever screwdriver
https://youtu.be/D36kpU-VaDA?feature=shared
I thought this was a clever invention. One end of the screwdriver shines a light on a screw and its camera determines the screw size, then flashes a signal beside the correct bit to use in the tray of bits, and sets the correct torque for the bit before it's inserted in the other end of the screwdriver handle. 🪛
Very clever screwdriver
https://youtu.be/D36kpU-VaDA?feature=shared
I thought this was a clever invention. One end of the screwdriver shines a light on a screw and its camera determines the screw size, then flashes a signal beside the correct bit to use in the tray of bits, and sets the correct torque for the bit before it's inserted in the other end of the screwdriver handle. 🪛
Wednesday, November 11, 2026
Monday, November 9, 2026
What makes streaming services so show to load
What Actually Causes the Load Time?
The delay you experience comes from technical constraints inherent to video streaming over the internet:
1. Manifest Retrieval & DRM Handshake: The TV app must request streaming manifests, authenticate user entitlement (subscription status), and establish digital rights management (DRM) decryption keys before playing protected content.
2. Buffer Building: Streaming platforms fetch several seconds of video into temporary memory (buffering) to ensure smooth playback even if network speeds fluctuate.
3. Adaptive Bitrate Selection: The app tests connection speed to select the optimal resolution stream (e.g., 1080p vs. 4K) before rendering the first frame.
4. Hardware & Hardware Decoding: Smart TV processors—especially on low-cost or built-in TV platforms—often have limited RAM and CPU resources, slowing down app execution and video decoding compared to dedicated streaming sticks or phones.
How to decrease 911 calls by 99%
https://youtu.be/ZoB91jIyG8Q?is=wcJ3FKZ0PDbs7ekI
People in crisis had nowhere to turn for simple services like helping someone get up off the floor, or getting a walker to prevent falls, that they were calling 911 almost 100 times a month. Instituting a social services division in their 911 system to follow up with callers dramatically decreased this overuse of 911
Sunday, November 8, 2026
Zapping weeds
https://youtu.be/PpnCu8SUGCU?is=CFxBe1FW5WZXCnDt
Zapping weeds one at a time can cut weeding costs per acre in half. Unfortunately, some hand weeding is still needed (within that reduced cost.)
Further reduced farming costs come from more efficient use of water and avoiding herbicides.
Healthier plants, healthier soil, healthier farmers.
This company is selling hundreds of units a year, planning to scale up to thousands.
% of population that are farmers
Early 1800s Agrarian Economy: Over 80% of workers were engaged in manual subsistence and small-scale commercial farming, requiring dense rural labor forces to feed the population.
1860–1900: The advent of steam-powered machinery, mechanical harvesters, and expanded rail distribution allowed fewer farmworkers to manage vastly larger crop yields.
1920–1960: Internal combustion tractors replaced draft animals, while synthetic fertilizers (Haber-Bosch process), hybrid seed stocks, and electrification augmented output per worker hour. Farm labor fell from 27% of the population in 1920 to under 7% by 1960.
Present: Commercial consolidation, GPS-guided precision agriculture, automated irrigation, and high-yield crop genetics reduced direct farm labor to approximately 1.3% of total workforce, even as total agricultural output is reaching historical highs.
Looking beyond USA, on a global scale, the percentage continues to decline and is now around 26%, as developing nations modernize their economies.
Subscribe to:
Posts (Atom)