It’s a bit ironic that an Atari 2600 game based on Raiders of the Lost Ark — a movie about archaeology — is now the subject of its own archaeological expedition as [Dennis Debro] and [Halkun] spent time reverse-engineering the game. Luckily, they shared their findings, so you can enjoy it the same way you can visit a king’s tomb without having to discover it and dig for it. If you don’t remember the game, you might enjoy the demo from [Speedy Walkthroughs] in the video below.

If you are only used to modern software, you might think this is little more than someone dumping the program code and commenting it. However, on these old, limited systems, you have to really understand the actual architecture because there are so many things you have to manage that are specific to the hardware.

For example, the game has two 4K ROM banks that use a strange switching mechanism. The entire game is built around the NTSC television signal. Everything is oriented toward generating the 60 Hz frame rate. Game logic runs during the vertical blanking and over-scan sections to prevent strange visible artifacts due to software running.

This is a fascinating look inside game coding as it existed around 1982. Of course, you can also run everything using emulation. Usually, our reverse engineering is more hardware-related. But we do love these old games, too.

youtube.com/embed/j3Py3T5J3e0?…

hackaday.com/2026/02/13/inside…

Cyberdecks are great projects, and [Salim Benbouziyane]’s scratch-built CM Deck is a fantastic specimen. It’s a clamshell-style cyberdeck with custom split keyboard, trackpad, optional external WiFi antenna, and some slick underlighting thanks to a translucent bottom shell. There’s even a hidden feature that seems super handy for a cyberdeck: a special USB-C port that, when plugged in to another host (like another computer), lets the cyberdeck act as an external keyboard and trackpad for that downstream machine.
The CM Deck is built around the Raspberry Pi Compute Module 5, which necessitates a custom PCB but offers more design freedom.
Notably, the CM Deck is custom-built around the Raspberry Pi Compute Model 5. When we first peeped the CM5 the small size was striking, but of course that comes at the cost of having no connectors, supporting hardware, or heat management. That’s something [Salim] embraced because it meant being able to put connectors exactly where he wanted them, and not have to work around existing hardware. A custom PCB let him to lay out his cyberdeck with greater freedom, less wasted space, and ultimately integrate a custom-built keyboard (with RP2040 and QMK firmware).

Even the final enclosure is custom-made, with 3D printing being used to validate the design and PCBway providing finished plastic shells in addition to manufacturing the PCBs. [Salim] admits that doing so was an indulgence, but his delight at the quality of the translucent purple undercarriage is palpable.

[Salim]’s video (embedded below) is a deep dive into the whole design and build process, and it’s a great watch for anyone interested in the kind of work and decisions that go into making something like this. Experienced folks can expect to nod in sympathy when [Salim] highlights gotchas like doing CAD work based on the screen’s drawings, only to discover later that the physical unit doesn’t quite match.

The GitHub repository contains the design files for everything, so give it a browse if you’re interested. [Salim] is no stranger to clean builds, so take a moment to admire his CRT-style Raspberry Pi terminal as well.

youtube.com/embed/U5GZeMm5nhI?…

Thanks [Keith Olson] for sharing the tip!

hackaday.com/2026/02/13/custom…

In the latest episode of the Hackaday Podcast, editors Elliot Williams and Tom Nardi start things off by discussing the game of lunar hide-and-seek that has researchers searching for the lost Luna 9 probe, and drop a few hints about the upcoming Hackaday Europe conference. From there they’ll marvel over a miniature operating system for the ESP32, examine the re-use of iPad displays, and find out about homebrew software development for an obscure Nintendo handheld. You’ll also hear about a gorgeous RGB 14-segment display, a robot that plays chess, and a custom 3D printed turntable for all your rotational needs. The episode wraps up with a sobering look at the dangers of industrial robotics, and some fascinating experiments to determine if a decade-old roll of PLA filament is worth keeping or not.

Check out the links below if you want to follow along, and as always, tell us what you think about this episode in the comments!

html5-player.libsyn.com/embed/…

Download this episode in DRM-free MP3 on your ESP32 with BreezyBox for maximum enjoyment.

Where to Follow Hackaday Podcast

Places to follow Hackaday podcasts:

• iTunes
• Spotify
• Stitcher
• RSS
• YouTube
• Check out our Libsyn landing page

News:

• Forget Waldo. Where’s Luna 9?
• Hints about Hackaday Europe 2026

What’s that Sound?

• Think you know that sound? Fill out the form for a chance to win!

Interesting Hacks of the Week:

• BreezyBox: A BusyBox-Like Shell And Virtual Terminal For ESP32
  
  • GitHub – valdanylchuk/breezyapps: ESP32 ELF apps for BreezyBox

  
  

• Lessons Learned After A Head-First Dive Into Hardware Manufacturing
• Upcycling An IPad Into A Touchscreen Display For Your PC
• Exploring Homebrew for the Pokémon Mini
  
  • Pokemon-Mini.net

  
  

• Designing A Compact RGB 14-Segment Display
• A Working Intercom From Antique Telephones

Quick Hacks:

• Elliot’s Picks:
  
  • ReMemory Is The Amnesia-hedging Buddy Backup You Didn’t Know You Needed
  • Building A Self-Playing Chess Board Robot
  • Why Haven’t Quantum Computers Factored 21 Yet?
    
    • [1301.7007] Pretending to factor large numbers on a quantum computer

    
    

  
  

• Tom’s Picks:
  
  • Gaming On An Arduino Uno Q In Linux
  • Scanning Table For The Professional Maker
  • Making A Hidden Door Status Sensor

  
  

Can’t-Miss Articles:

• How Industrial Robot Safety Was Written In Blood
  
  • U.S. Chemical Safety Board

  
  

• Is That Ancient Reel Of PLA Any Good?

hackaday.com/2026/02/13/hackad…

One of the best parts of 3D printing is that you can freely download the plans for countless model from sites like Thingiverse, Printables, and others. Yet with the veritable flood of models on these sites you also want to have some level of quality. Here recent news pertaining to Thingiverse is probably rather joyful, as with the acquisition of Thingiverse by MyMiniFactory, it should remain one of the most friendly sites for sharing 3D printing models.

Although Thingiverse as a concept probably doesn’t need much introduction, it’s important here to acknowledge the tumultuous times that it has gone through since its launch in 2008 as part of MakerBot. Both were acquired by Stratasys in 2013, and this has lead to ups and downs in the relationship with Thingiverse’s user base.

MyMiniFactory was launched in 2013 as a similar kind of 3D printing object-sharing platform as Thingiverse, while also offering crowdsourcing and paid model options. In the MyMiniFactory blog post it’s stated that these features will not be added to Thingiverse, and that nothing should change for Thingiverse users in this regard.

What does change is its joining of the ‘SoulCrafted‘ initiative, which is an initiative against machine-generated content, including so-called ‘AI slop’. There will be a live Q & A on February 17th during which the community can pitch their questions and ideas, along with a dedicated Thingiverse group.

hackaday.com/2026/02/13/mymini…

We live in an age where engineering marvels are commonplace: airplanes crisscross the sky, skyscrapers grow like weeds, and spacecraft reach for the stars. But every so often, we see something unusual that makes us take a second look. The Falkirk Wheel is a great example, and, even better, it is functional art, as well.

The Wheel links two canals in Scotland. Before you click away, here’s the kicker: One canal is 35 meters higher than the other. Before 1933, the canals were connected with 11 locks. It took nearly a day to operate the locks to get a boat from one canal to the other. In the 1930s, there wasn’t enough traffic to maintain the locks, and they tore them out.

Fast Forward

In the 1990s, a team of architects led by [Tony Kettle] proposed building a wheel to transfer boats between the two canals. The original model was made from [Tony’s] daughter’s Lego bricks.

The idea is simple. Build a 35-meter wheel with two cassions, 180 degrees apart. Each cassion can hold 250,000 liters of water. To move a boat, you fill the caissons with 500 tonnes of water. Then you let a boat into one of them with its weight displacing an equal amount of water, so the caissons stay at the same weight.

Once you have a balanced system, you just spin the wheel to make a half turn. There are 10 motors that require 22.5 kilowatts, and each half turn consumes about 1.5 kilowatt-hours.

Not Lockless

The wheel actually raises boats up 24 m, so the remaining 11 m still requires two locks. But this is a far cry from the eleven locks the system replaces. The structure has a foundation with 30 concrete piles down on the bedrock. The wheel itself uses 14,000 bolts to avoid welds that might fatigue under stress.

As you’d expect, the caissons have to turn with the wheel in order to stay level, somewhat like a Ferris Wheel. This works using three 8-meter gears. It takes about four minutes for the wheel to make a half turn. You can watch it work in the video below.

Why?

We were a bit disappointed that there doesn’t seem to be any reason to connect the two canals except as a tourist attraction. On the other hand, about half a million visitors go every year, so it does have an economic impact. As far as we know, this is the world’s only rotating boat lift. It certainly is artistic compared to, say, the historic Anderton Lift.

We love big engineering. Even the ones that seem commonplace.

youtube.com/embed/qHO9gARac-w?…

Featured image: “FalkirkWheelSide” by Sean Mack.

hackaday.com/2026/02/13/the-en…

«Qualcosa è andato storto. Come i social network e l’intelligenza artificiale ci hanno rubato il futuro», è un potente atto d’accusa nei confronti della tecnologia che ci rende dipendenti da smartphone, tablet e computer. Pubblicato da Solferino pochi mesi fa, è già un caso editoriale. Alla settima ristampa ha venduto circa 20.000 copie. Un successo, per un paese dove un italiano su due legge meno di un libro all’anno.

E il motivo di questo successo è che il libro spiega, in parole chiare e comprensibili, senza peli sulla lingua, che cosa le aziende come Meta/Facebook e TikTok, sono state e sono pronte a fare pur di catturarci davanti allo schermo, per vendere il nostro tempo di attenzione agli inserzionisti e così facendo terremotando psiche, sentimenti e percezioni degli utenti, soprattutto di quelli più giovani.

Già, perché, Riccardo Luna, è lui l’autore del libro, ripercorrendo le origini della Rete, che lui stesso da direttore di Wired candidò a premio nobel per la pace, racconta come questa si sia ormai popolata di app che recintano i nostri interessi e argomenta come la Rete, da utopia democratica dei primi anni 2000 sia stata colonizzata dall’affarismo di Big Tech che non ritiene di dover rispondere a nessuno del proprio operato. Aziende guidate da imprenditori tech che anziché dedicare le loro enormi ricchezze a sfamare il mondo, a contrastare gli effetti perversi del cambiamento climatico e a favorire pace e democrazia, lucrano su sistemi che creano rabbia e violenza, i social network a cui hanno messo il turbo degli algoritmi capaci di innescare i circuiti del piacere che scattano quando vediamo il like sotto il nostro post e ci rendono frustrati quando non li vediamo. Meccanismi che sfruttano anche la solitudine, l’incertezza sessuale e le delusioni dei più giovani per renderli sempre meno sicuri di se stessi e sempre più dipendenti dal feed del social preferito.

Temi già sollevati da altri prima di lui, certamente, come Nicholas Carr, Jaron Lanier, l’italiano Jacopo Franchi. Autori che hanno raccontato gli effetti dannosi dello scrolling infinito, quel modo di usare i social senza smettere mai e che, lungi dall’essere una cosa naturale innescata dalla curiosità, è un comportamento ingegnerizzato per le piattaforme da tecnologi e psicologi sociali in modo da trattenere chiunque, sia esso bambino o un adulto, appiccicati allo smartphone grazie all’algoritmo di selezione dei contenuti. A proposito, tu che leggi lo sapevi che basta un centinaio di video visti per riorganizzare l’algoritmo?

«Qualcosa è andato storto. Come i social network e l’intelligenza artificiale ci hanno rubato il futuro», Riccardo Luna, Solferino, 2025.

dicorinto.it/articoli/recensio…

Pneumatics are a common way to add some motion to soft robotic actuators, but adding it to a robot can be somewhat of a chore. A method demonstrated by [Jackson K. Wilt] et al. (press release, preprint) involves using a 3D printing to extrude two materials: one elastomeric material and a fugitive ink that is used to create pneumatic channels which are dissolved after printing, leaving the empty channels to be filled with air.

By printing these materials with a rational, multi-material (RM-3DP) custom nozzle it’s possible to create various channel patterns, controlling the effect of compressed air on the elastomeric material. This way structures like hinges and muscles can be created, which can then be combined into more complex designs. One demonstrated design involves a human-like hand with digits that can move and grasp, for example.

In the demonstration the elastomeric material is photopolymerizable polyurethane-acrylate resin, with the fugitive ink being 30 wt% Pluronic F-127 in water. The desired pattern is determined beforehand with a simulation, followed by the printing and UV curing of the elastomeric resin.

As is typical of soft robotics implementations, the resulting robots are more about a soft touch than a lot of force, but could make for interesting artificial muscle designs due to how customizable the printing process is.

youtube.com/embed/BK9K_mJjlxE?…

hackaday.com/2026/02/13/3d-pri…

The classic design of a mechanical clock generally consists of a display, a way to store energy, a way to release that energy at regular intervals, and a mechanism to transmit it where it needs to go. Most of us might be imagining a pendulum or a balance wheel, but there have been many other ways to maintain a reliable time standard with a physical object beyond these two common methods. This clock, for example, uses a rolling ball bearing as its time standard and [Tommy Jobson] discusses its operation in depth during a restoration.

The restoration of this clock, which [Tommy] theorizes was an amateur horological project even when it was new, starts by dismantling the clock nearly completely. The clock was quite dirty, so in addition to being thoroughly cleaned it also needed a bit of repair especially involving a few bent pins that stop the table’s rotation. These pins were replaced with stronger ones, and then everything in the clock’s movement was put back together. The tray carrying the ball bearing needed to be cleaned as well, and [Tommy] also added a lacquer to help preserve the original finish as long as possible. From there it was time to start calibrating the clock.

The ball bearing itself rolls back and forth along an inclined plane on a series of tracks. When it gets to the end it hits a lever which lets a bit of energy out of the movement, tilting the table back in the other direction to repeat the process. This is a much more involved process for getting an accurate time interval than a pendulum, so [Tommy] had a lot of work to do here. But in the end he was able to bring it back to life with an accuracy fairly close to a pendulum clock.

Ball bearings are a pretty popular medium for clock builds even in the modern era. This one uses them in a unique display, and a more recent version goes even further by using marbles to display digits directly.

youtube.com/embed/_6IgX2WxFDI?…

Thanks to [Keith] for the tip!

hackaday.com/2026/02/13/restor…

The Argon40 ONE UP unsurprisingly looks like a laptop. (Credit: Jeff Geerling)
The Raspberry Pi Compute Module form factor is a tantalizing core for a potential laptop, with a CM5 module containing a fairly beefy SoC and RAM, with depending on the exact module also eMMC storage and WiFi. To turn this into a laptop you need a PCB to put the CM5 module on and slide it into a laptop shell. This is in effect what [Argon40] did with their crowdfunded ONE UP laptop, which [Jeff Geerling] has been tinkering with for a few weeks now, with some thoughts on how practical the concept of a CM5-based laptop is.

Most practical is probably the DIY option that [Jeff] opted for with the ‘Shell’ version that he bought, as that meant that he could pop in one of the CM5s that he had lying around. The resulting device is totally functional as a laptop, with all the Raspberry Pi 5 levels of performance you’d expect and with the repair-friendliness of a Framework laptop.

If you’re buying the Core version with the 8 GB CM5 module and 256 GB NVMe SSD included, you’re looking at €475 before shipping or the equivalent in your local currency. This puts it unfortunately in the territory of budget x86 laptops and used Apple MacBooks, even before taking into account the current AI-induced RAMpocalypse that’d push [Jeff]’s configuration to $600 if purchased new, with prices likely to only go up.

Even if this price isn’t a concern, and you just want to have a CM5-based laptop, [Jeff]’s experience got soured on poor customer support from [Argon40] and above all the Raspberry Pi’s arch nemesis: the inability to do sleep mode. With the lid closed it runs at 3.3 W idle, but that’ll run down the battery from 100% to flat in about 17 hours. Perhaps if Raspberry Pi added sleep states to their systems would it make for a good laptop core, as well as for a smartphone.

youtube.com/embed/Ef70x0izkFU?…

hackaday.com/2026/02/12/argon-…

[The Signal Path] shows us how to recreate a classic science experiment to measure the weight of an electron. Things are easier for us, because unlike [J. J. Thomson] in 1897, we have ready sources of electrons and measuring equipment. Check it out in the video below.

The main idea is to trap an electron using a magnetic field into a circular path. You can then compute the forces required to keep it in that circle, along with some other equations, and combine them. The result lets you compute the charge to mass ratio using parameters you can either control or measure, like the radius of the circular path and the electric field.

Helmholtz coils create the magnetic field, and a cold cathode tube provides the electrons. Honestly, the equipment looks a bit like something out of an old monster movie.

Of course, the result is the charge to mass ratio, which means to get the mass, you need to know the charge of the electron. Today, you can look that up, but in 1897, no one knew what it was. [Robert Millikan] would conduct another experiment using oil drops about a decade later to determine that number, and then the world could know the mass of a single electron.

The resulting ratio was very close to the accepted value. It would be fun to see someone replicate the oil drop experiment, too. You could spend a lot of time recreating classic science experiments. Some of the experiments are easy with today’s gear.

youtube.com/embed/AI5NuiIAdCY?…

hackaday.com/2026/02/12/does-t…

For as long as small, hidden radio transmitters have existed, people have wanted a technology to detect them. One of the more effective ways to find hidden electronics is the nonlinear junction detector, which illuminates the area under investigation with high-frequency radio waves. Any P-N semiconductor junctions in the area will emit radio waves at harmonic frequencies of the original wave, due to their non-linear electronic response. If, however, you suspect that the electronics might be connected to a dangerous device, you’ll want a way to detect them from a distance. One solution is harmonic radar (also known as nonlinear radar), such as this phased-array system, which detects and localizes the harmonic response to a radio wave.

One basic problem is that semiconductor devices are very rarely connected to antennas optimized for the transmission of whatever harmonic you’re looking for, so the amount of electromagnetic radiation they emit is extremely low. To generate a detectable signal, a high-power transmitter and a very high-gain receiver are necessary. Since semiconductor junctions emit stronger lower harmonics, this system transmits in the 3-3.2 GHz range and only receives the 6-6.4 GHz second harmonic; to avoid false positives, the transmitter provides 28.8 decibels of self-generated harmonic suppression. To localize a stronger illumination signal to a particular point, both the transmit and receive channels use beam-steering antenna arrays.

In testing, the system was able to easily detect several cameras, an infrared sensor, a drone, a walkie-talkie, and a touch sensor, all while they were completely unpowered, at a range up to about ten meters. Concealing the devices in a desk drawer increased the ranging error, but only by about ten percent. Even in the worst-case scenario, when the system was detecting multiple devices in the same scene, the ranging error never got worse than about 0.7 meters, and the angular error was never worse than about one degree.

For a refresher on the principles of the technology, we’ve covered nonlinear junction detectors before. While the complexity of this system seems to put it beyond the reach of amateurs, we’ve seen some equally impressive homemade radar systems before.

hackaday.com/2026/02/12/harmon…

Ham radio operators may be familiar with slow-scan television (SSTV) where an image is sent out over the airwaves to be received, decoded, and displayed on a computer monitor by other radio operators. It’s a niche mode that isn’t as popular as modern digital modes like FT8, but it still has its proponents. SSTV isn’t only confined to the radio, though. [BLANCHARD Jordan] used this encoding method to store digital images on a cassette tape in a custom-built tape deck for future playback and viewing.

The self-contained device first uses an ESP32 and its associated camera module to take a picture, with a screen that shows the current view of the camera as the picture is being taken. In this way it’s fairly similar to any semi-modern digital camera. From there, though, it starts to diverge from a typical digital camera. The digital image is converted first to analog and then stored as audio on a standard cassette tape, which is included in the module in lieu of something like an SD card.

To view the saved images, the tape is played back and the audio signal captured by an RP2040. It employs a number of methods to ensure that the reconstructed image is faithful to the original, but the final image displays the classic SSTV look that these images tend to have as a result of the analog media. As a bonus feature, the camera can use a serial connection to another computer to offload this final processing step.

We’ve been seeing a number of digital-to-analog projects lately, and whether that’s as a result of nostalgia for the 80s and 90s, as pushback against an increasingly invasive digital world, or simply an ongoing trend in the maker space, we’re here for it. Some of our favorites are this tape deck that streams from a Bluetooth source, applying that classic cassette sound, and this musical instrument which uses a cassette tape to generate all of its sounds.

hackaday.com/2026/02/12/storin…

Originally only sold at the Pokémon Center New York in late 2001 for (inflation adjusted) $80, the Pokémon Mini would go on to see a release in Japan and Europe, but never had more than ten games produced for it. Rather than Game Boy-like titles, these were distinct mini games that came on similarly diminutive cartridges. These days it’s barely remembered, but it can readily be used for homebrew titles, as [Inkbox] demonstrates in a recent video.

Inside the device is an Epson-manufactured 16-bit S1C88 processor that runs at 4 MHz and handles basically everything, including video output to the monochrome 96×64 pixel display. System RAM is 4 kB of SRAM, which is enough for the basic games that it was designed for.

The little handheld system offered up some capabilities that even the full-sized Game Boy couldn’t match, such as a basic motion sensor in the form of a reed relay. There’s also 2 MB of ROM space directly addressable without banking.

Programming the device is quite straightforward, not only because of the very accessible ISA, but also the readily available documentation and toolchain. This enables development in C, but in the video assembly is used for the added challenge.

Making the screen tiles can be done in an online editor that [Inkbox] also made, and the game tested in an emulator prior to creating a custom cartridge that uses an RP2040-based board to play the game on real hardware. Although a fairly obscure gaming handheld, it seems like a delightful little system to tinker with and make more games for.

youtube.com/embed/48Mg4YMJGIk?…

hackaday.com/2026/02/12/explor…

Anyone who has spent any amount of time in or near people who are really interested in energy policies will have heard proclamations such as that ‘baseload is dead’ and the sorting of energy sources by parameters like their levelized cost of energy (LCoE) and merit order. Another thing that one may have noticed here is that this is also an area where debates and arguments can get pretty heated.

The confusing thing is that depending on where you look, you will find wildly different claims. This raises many questions, not only about where the actual truth lies, but also about the fundamentals. Within a statement such as that ‘baseload is dead’ there lie a lot of unanswered questions, such as what baseload actually is, and why it has to die.

Upon exploring these topics we quickly drown in terms like ‘load-following’ and ‘dispatchable power’, all of which are part of a healthy grid, but which to the average person sound as logical and easy to follow as a discussion on stock trading, with a similar level of mysticism. Let’s fix that.

Loading The Bases

Baseload is the lowest continuously expected demand, which sets the minimum required amount of power generating capacity that needs to be always online and powering the grid. Hence the ‘base’ part, and thus clearly not something that can be ‘dead’, since this base demand is still there.

What the claim of ‘baseload is dead’ comes from is the idea that with new types of generation that we are adding today, we do not need special baseload generators any more. After all, if the entire grid and the connected generators can respond dynamically to any demand change, then you do not need to keep special baseload plants around, as they have become obsolete.
Example electrical demand “Duck Curve” using historical data from California. (Credit: ArnoldRheinhold)
A baseload plant is what is what we traditionally call power plants that are designed to run at 100% output or close to it for as long as they can, usually between refueling and/or maintenance cycles. These are generally thermal plants, powered by coal or nuclear fuel, as this makes the most economical use of their generating capacity, and thus for the cheapest form of dispatchable power on the grid.

With only dispatchable generators on the grid this was very predictable, with any peaks handled by dedicated power plants, both load-following and peaking power plants. This all changed when large-scale solar and wind generators were introduced, and with it the duck curve was born.

As both the sun and wind are generally more prevalent during the day, and these generators are not generally curtailed, this means that suddenly everything else, from thermal power plants to hydroelectric plants, has to throttle back. Obviously, doing so ruins the economics of these dispatchable power sources, but is a big part of why the distorted claim of ‘baseload is dead’ is being made.

Chaos Management

The Fengning pumped storage power station in north China’s Hebei Province. (Credit: CFP)
Suffice it to say that having the entire grid adapt to PV solar and wind farms – whose output can and will fluctuate strongly over the course of the day – is not an incredibly great plan if the goal is to keep grid costs low. Not only can these forms of variable renewable energy (VRE) only be curtailed, and not ramped up, they also add thousands of kilometers of transmission lines and substations to the grid due to the often remote areas where they are installed, adding to the headache of grid management.

Although curtailing VRE has become increasingly more common, this inability to be dispatched is a threat to the stability of the national grids of countries that have focused primarily on VRE build-out, not only due to general variability in output, but also because of “anticyclonic gloom“: times when poor solar conditions are accompanied by a lack of wind for days on end, also called ‘Dunkelflaute’ if you prefer a more German flair.

What we realistically need are generators that are dispatchable – i.e. are available on demand – and can follow the demand – i.e. the load – as quickly as possible, ideally in the same generator. Basically the grid controller has to always have more capacity that can be put online within N seconds/minutes, and have spare online capacity that can ramp up to deal with any rapid spikes.

Although a lot is being made of grid-level storage that can soak up excess VRE power and release it during periods of high demand, there is no economical form of such storage that can also scale sufficiently. Thus countries like Germany end up paying surrounding countries to accept their excess power, even if they could technically turn all of their valleys into pumped hydro installations for energy storage.

This makes it incredibly hard to integrate VRE into an electrical grid without simply hard curtailing them whenever they cut into online dispatchable capacity.

Following Dispatch

Essential to the health of a grid is the ability to respond to changes in demand. This is where we find the concept of load-following, which also includes dispatchable capacity. At its core this means a power generator that – when pinged by the grid controller (transmission system operator, or TSO) – is able to spin up or down its power output. For each generator the response time and adjustment curve is known by the TSO, so that this factor can be taken into account.
European-wide grid oscillations prior to the Iberian peninsula blackout. (Credit: Linnert et al., FAU, 2025)
The failure of generators to respond as expected, or by suddenly dropping their output levels can have disastrous effects, particularly on the frequency and thus voltage of the grid. During the 2025 Iberian peninsula blackout, for example, grid oscillations caused by PV solar farms caused oscillation problems until a substation tripped, presumably due to low voltage, and a cascade failure subsequently rippled through the grid. A big reason for this is the inability of current VRE generators to generate or absorb reactive power, an issue that could be fixed with so-called grid-forming converters, but at significant extra cost to the VRE generator owners, as this would add local energy storage requirements such as batteries.

Typically generators are divided into types that prefer to run at full output (baseload), can efficiently adjust their output (load follow) or are only meant for times when demand outstrips the currently available supply (peaker). Whether a generator is suitable for any such task largely depends on the design and usage.

This is where for example a nuclear plant is more ideal than a coal plant or gas turbine, as having either of these idling burns a lot of fuel with nothing to show for it, whereas running at full output is efficient for a coal plant, but is rather expensive for a gas turbine, making them mostly suitable for load-following and peaker plants as they can ramp up fairly quickly.

The nuclear plant on the other hand can be designed in a number of ways, making it optimized for full output, or capable of load-following, as is the case in nuclear-heavy countries like France where its pressurized water reactors (PWRs) use so-called ‘grey control rods’ to finely tune the reactor output and thus provide very rapid and precise load-following capacities.
Overview of the thermal energy transfer in the Natrium reactor design. (Source: TerraPower)
There’s now also a new category of nuclear plant designs that decouple the reactor from the steam turbine, by using intermediate thermal storage. The Terrapower Natrium reactor design – currently under construction – uses molten salt for its coolant, and also molten salt for the secondary (non-nuclear) loop, allowing this thermal energy to be used on-demand instead of directly feeding into a steam turbine.

This kind of design theoretically allows for a very rapid load-following, while giving the connected reactor all the time in the world to ramp up or down its output, or even power down for a refueling cycle, limited only by how fast the thermal energy can be converted into electrical power, or used for e.g. district heating or industrial heat.

Although grid-level storage in the form of pumped hydro is very efficient for buffering power, it cannot be used in many locations, and alternatives like batteries are too expensive to be used for anything more than smoothing out rapid surges in demand. All of which reinforces the case for much cheaper and versatile dispatchable power generators.

Grid Integration

Any power generator on the grid cannot be treated as a stand-alone unit, as each kind of generator comes with its own implications for the grid. This is a fact that is conveniently ignored when the so-called Levelized Cost of Energy (LCoE) metric is used to call VRE the ‘cheapest’ of all types of generators. Although it is true that VRE have no fuel costs, and relatively low maintenance cost, the problem with them is that most of their costs is not captured in the LCoE metric.

What LCoE doesn’t capture is whether it’s dispatchable or not, as a dispatchable generator will be needed when a non-dispatchable generator cannot produce due to clouds, night, heavy snow cover, no wind or overly strong wind. Also not captured in LCoE are the additional costs occurred from having the generator connected to the grid, from having to run and maintain transmission lines to remote locations, to the cost of adjusting for grid frequency oscillations and similar.
Levelized cost of operation of various technologies. (Credit: IEA, 2020)
Ultimately these can be summarized as ‘system integration costs’, and they are significantly tougher to firmly nail down, as well as highly variable depending on the grid, the power mix and other variables. Correspondingly the cost of electricity from various sources is hotly debated, but the consensus is to use either Levelized Avoided Cost of Energy (LACE) or Value Adjusted LCoE (VALCoE), which do take these external factors into account.
Energy value by technology relative to average wholesale electricity price in the European Union in the Stated Policies Scenario. (Credit: IEA, 2020)
As addressed in the linked IEA article on VALCoE, an implication of this is that the value of VREs drop as their presence on the grid increases. This can be seen in the above graph based on 2020-era EU energy policies, with the graphs for the US and China being different again, but China’s also showing the strong drop in value of PV solar while wind power is equally less affected.

A Heated Subject

It is unfortunate that energy policy has become a subject of heated political and ideological furore, as it should really be just as boring as any other administrative task. Although the power industry has largely tried to stay objective in this matter, it is unfortunately subject to both political influence and those of investors. This has led to pretty amazing and breakneck shifts in energy policy in recent years, such as Belgium’s phase-out of nuclear power, replacing it with multiple gas plants, to then not only decide to not phase out its existing nuclear plants, but also to look at building new nuclear.

Similarly, the US has and continues to see heated debates on energy policy which occasionally touch upon objective truth. Unfortunately for all of those involved, power grids do not care about personal opinions or preferences, and picking the wrong energy policy will inevitably lead to consequences that can cost lives.

In that sense, it is very harmful that corner stones of a healthy grid such as baseload, reactive power handling and load-following are being chipped away by limited metrics such as LCoE and strong opinions on certain types of power technologies. If we cared about a stable grid more than about ‘being right’, then all VRE generators would for example be required to use grid-forming converters, and TSOs could finally breathe a sigh of relief.

hackaday.com/2026/02/12/the-de…

It is no secret that we often use and abuse bash to write things that ought to be in a different language. But bash does have its attractions. In the modern world, it is practically everywhere. It can also be very expressive, but perhaps hard to read.

We’ve talked about Amber before, a language that is made to be easier to read and write, but transpiles to bash so it can run anywhere. The FOSDEM 2026 conference featured a paper by [Daniele Scasciafratte] that shows how to best use Amber. If you prefer slides to a video, you can read a copy of the presentation.

For an example, here’s a typical Amber script. It compiles fully to a somewhat longer bash script:

import * from "std/env"
fun example(value:Num = 1) {
if 1 > 0 {
let numbers = [value, value]
let sum = 0
loop i in numbers {
sum += numbers
[i] }
echo "it's " + "me"
return sum
}
fail 1
}

echo example(1) failed {
echo "What???"
is_command("echo")
}

The slides have even more examples. The language seems somewhat Python-like, and you can easily figure out most of it from reading the examples. While bash is nearly universal, the programs a script might use may not be. If you have it, the Amber code will employ bshchk to check dependencies before execution.

According to the slides, zsh support is on the way, too. Overall, it looks like it would be a great tool if you have to deploy with bash or even if you just want an easier way to script.

We’ve looked at Amber before. Besides, there are a ton of crazy things you can do with bash.

hackaday.com/2026/02/12/bash-v…