If you talk to the FDA, there’s only one permanent method of hair removal—electrolysis. This involves sticking a needle into a hair follicle, getting it very hot or running a current through it, and then letting heat and/or the lye generated kill the root of the hair dead. Normally, you’d pay someone with a commercial machine to do this for you at great expense. Or, you could do it yourself with a home-built machine, as [n3tcat] did.

Based on the available information out in the wild, [n3tcat] decided to build a galvanic electrolysis machine. This specifically passes current through a needle in the hair follicle to generate lye at the hair bulb, which kills it. The amount of lye generated depends on the amount of current and the time over which it is applied. More lye is more likely to kill a follicle permanently, though there are limits with regards to avoiding scarring, other skin damage, and excessive pain.

[n3tcat]’s guide explains the basic theory behind galvanic electrolysis, as well as how the rig was built. An early attempt simply involved hooking up a 12-volt car battery to a standard electrolysis needle, sticking it in a hair, with the other electrode being an aluminium can held by the person being treated. The fun thing was that this allowed varying the current depending on how much contact and how stiffly the person grabbed the can.

After a few successful hair removals this way, [n3tcat] decided to build a better rig. An RP2040 microcontroller was enlisted to run the show, powered by a 3.7-volt lithium rechargeable battery. An OLED screen and a rotary encoder were selected to serve as the interface, while a foot pedal was added for firing off current. A boost converter was used to push the battery voltage up to the vicinity of 15 volts for delivery to the needle, set up to avoid excessive current delivery for safety. A DAC was paired with an LM358 op-amp feeding into a MOSFET to control the current passed to the needle for accurate, controlled treatment, with the RP2040 monitoring the current level via a dedicated ADC. The needle itself got a D-printed pen-like handle for better ergonomics, easing the process of slotting the needle into a hair follicle. Everything was then assembled on a cute PCB, and wrapped up in a nice 3D printed housing. The files are available for the curious.

Electrolysis is a process that can cost many thousands of dollars depending on how much hair you hope to remove. Thus, it’s easy to see the appeal in having a rig that lets you do it at home. It’s just one of those things where you have to take the proper precautions to ensure you’re not unduly hurting yourself.  Stay safe out there, hackers!

Even as leading AI providers like OpenAI and Anthropic battle over the compute to train and release ever larger, more powerful models, other labs are going in a different direction — pursuing the development of smaller, more efficient models and often open sourcing them.

The latest worth paying attention to comes from the lesser-known Palo Alto startup Zyphra, which this week released its new reasoning, mixture-of-experts (MoE) language model, ZAYA1-8B, with just over 8 billion parameters and only 760 million active — far fewer than the trillions estimated for the likes of the big labs. Yet, ZAYA1-8B retains competitive performance on third-party benchmarks against GPT-5-High and DeepSeek-V3.2.

It can be downloaded from Hugging Face now free of charge under a permissive, standard, enterprise-friendly Apache 2.0 license — and enterprises and indie developers can begin using and customizing it immediately to suit their needs. Individual users can also test it themselves here free at Zyphra Cloud, the startup’s inference solution.

But the real headline is what ZAYA1-8B was trained on: a full stack of AMD Instinct MI300 graphics processing units (GPUs), the rival to Nvidia GPUs released by AMD nearly three years ago, and which shows that this platform is capable of producing useful models and is a viable alternative to the preferential position Nvidia has maintained in recent years among AI model developers.

How ZAYA1-8B was trained

The “intelligence density” touted by Zyphra is the result of what they describe as a “full-stack innovation” approach, spanning architecture, pretraining, and reinforcement learning (RL).

ZAYA1-8B is built on Zyphra’s proprietary MoE++ architecture, described in a technical report released by the lab. This architecture introduces three fundamental changes to the standard Transformer architecture that gave rise to large language models (LLMs) and the entire generative AI era:

1. Compressed Convolutional Attention (CCA): Unlike standard attention mechanisms that struggle with memory as context windows grow, CCA performs sequence mixing in a compressed latent space. This results in an 8x reduction in KV-cache size compared to full multi-head attention, enabling more efficient long-context reasoning.

2. The ZAYA1 MLP Router: Most MoE models use a linear router to decide which “experts” handle a specific token. Zyphra replaced this with a more expressive multi-layer MLP-based design. To maintain stability during training—a common hurdle for MoEs—they implemented a bias-balancing scheme inspired by PID controllers from classical control theory.

3. Learned Residual Scaling: This controls the growth of the “residual norm” as data flows deeper into the model’s 40 layers, preventing gradient vanishing or explosion with negligible computational overhead.

Reasoning-First Pretraining

A critical differentiator for ZAYA1-8B is that reasoning was integrated from the start of pretraining, rather than being “bolted on” during post-training.

To handle long chain-of-thought (CoT) traces that would otherwise exceed the initial 4K pretraining context, Zyphra developed Answer-Preserving (AP) Trimming.

Think of AP-trimming like a film editor cutting a long scene: instead of cutting the ending (the solution) or dropping the scene entirely, the editor removes the “middle” of the character’s monologue while keeping the beginning (the problem setup) and the final reveal (the answer).

This ensures the model learns the relationship between complex problems and their solutions even when the full internal logic doesn’t yet fit into memory.

It seemed to work well on my test query about countertop stain removal to ZAYA1-8B running on Zyphra Cloud.

Markovian RSA: redefining test-time compute

The model’s most significant performance leap comes from Markovian RSA, a novel test-time compute (TTC) methodology.

Traditionally, if you want a model to “think harder,” you let it generate a longer chain of thought. However, this often leads to “context bloat,” where the model loses focus as the history grows too long.

Markovian RSA solves this by decoupling “thinking depth” from “context size”. It functions like a recursive scientific peer-review process:

• The model generates multiple parallel reasoning traces (candidates).

• It then extracts only the “tails” (the last few thousand tokens) of these traces.

• These tails are subsampled and presented to the model in a new “aggregation prompt,” asking it to reconcile the different approaches into a better solution.

By carrying forward only the tails (typically a 4K-token budget), the model can reason indefinitely without the context window ever overflowing. In practice, this allows the 700M active parameter ZAYA1-8B to achieve a 91.9% score on AIME ’25, closing the gap with models that have 30 to 50 times its active parameter count.

Because ZAYA1-8B maintains a small total parameter footprint (8.4B), it is uniquely positioned for on-device deployment and local LLM applications. For enterprises, this enables the deployment of high-tier reasoning capabilities—traditionally reserved for massive cloud-based models—directly onto local hardware or edge devices. This “local-first” reasoning approach addresses common enterprise hurdles regarding data residency, latency, and the high cost of persistent API dependencies.

Benchmarks show a remarkably performant small model that punches above its weight class

Zyphra is positioning ZAYA1-8B as a “punch above its weight” model for developers who need high-tier reasoning without the latency or cost of massive frontier models. After all, its active parameter count is much lower than other similarly-sized models, making it much cheaper and less compute-intensive to run in inference.

• Instruction Following: ZAYA1-8B scores 85.58 on IFEval, remaining competitive with much larger models like Intellect-3 (106B).

• Agentic Capabilities: On the τ² benchmark, the model reaches 43.12, and 39.22 on BFCL-v4, providing a baseline for its ability to handle tool-calling and multi-turn tasks.

In single-rollout evaluations (without the extra “thinking” time), ZAYA1-8B already outperforms its weight class. It beats Qwen3.5-4B and Gemma-4-E4B on math and code benchmarks.

When Markovian RSA is enabled, the results are startling:

• HMMT ’25 (Math): ZAYA1-8B hits 89.6%, surpassing Claude 4.5 Sonnet (79.2%) and GPT-5-High (88.3%).

• LiveCodeBench (Coding): The model achieves 69.2%, outperforming DeepSeek-R1-0528.

Zyphra notes that while the model is a specialist in algorithmic reasoning, it lags slightly behind larger models on “knowledge-heavy” tasks like broad factual retrieval (MMLU-Pro), which suggests that while reasoning can be compressed into smaller cores, factual memory still benefits from raw parameter count.

Apache 2.0 open licensed for research and commercial usage

Zyphra has released ZAYA1-8B under the Apache-2.0 license. This is a critical choice for the developer community. Unlike “copyleft” licenses like the GPL, which require any derived work to also be open-source, Apache-2.0 is highly permissive.

For developers and enterprises, this means they can use, modify, and distribute ZAYA1-8B—even within proprietary, commercial applications—without being forced to open-source their own codebases.

It also includes an explicit grant of patent rights from contributors, providing a layer of legal safety for startups building on top of Zyphra’s architecture. By opting for Apache-2.0 over more restrictive “research-only” licenses often seen from frontier labs, Zyphra is signaling a commitment to the open-weight ecosystem.

To deploy ZAYA1-8B, developers must use specific branches from Zyphra’s forks of core libraries, as the architecture requires specialized handling:

• Custom Forks: Users should install the zaya1 branch from Zyphra’s versions of the vllm and transformers libraries.

• Deployment Flags: When starting a vLLM server, specific flags are required to handle the reasoning parser and tool-calling (e.g., --reasoning-parser qwen3 and --tool-call-parser zaya_xml).

• Parallelism Strategy: For multi-GPU environments, Zyphra recommends using Data Parallelism (DP) combined with Expert Parallelism (EP). Notably, Tensor Parallelism (TP) for the model’s CCA mechanism is not currently supported, making DP+EP the optimal path for scaling inference throughput.

Background on Zyphra

Zyphra: A New Paradigm for Intelligence Density

Founded in 2021 and headquartered in Palo Alto, California, Zyphra Technologies is a full-stack artificial intelligence laboratory dedicated to building human-aligned artificial general intelligence (AGI) — that which outperforms people at most tasks — through a decentralized, open-source framework.

According to the company’s official mission statement, Zyphra seeks to challenge the “centralized” dominance of monolithic cloud models by focusing on “intelligence density”—a core guiding principle that aims to maximize the reasoning and logic extracted per parameter and per FLOP.

Zyphra CEO and Co-Founder Krithik Puthalath explained previously to VentureBeat that this strategy is essential for enabling high-performance AI to run locally on hardware such as tablets, wearable glasses, and enterprise servers, thereby ensuring user privacy and reducing reliance on third-party cloud infrastructure.

The company’s technical identity is deeply informed by computational neuroscience, led by Co-Founder and Chief Scientist Beren Millidge.

According to Millidge’s personal website, he currently serves as a Postdoctoral Researcher at the University of Oxford’s Nuffield Department of Clinical Neurosciences, where his research focuses on deep credit assignment and mathematical models of the brain.

Millidge, who earned his PhD from the University of Edinburgh, has pioneered research into active inference and the “free-energy principle,” concepts that directly influence Zyphra’s pursuit of multimodal architectures capable of long-term memory and continual learning.

This neuroscientific influence was central to the design of Zyphra’s prior Zamba model, released in 2024, which mimics the cortex-hippocampus interaction to share information across sequential layers. A recent TED Talk video provides insight into Millidge’s perspective on the intersection of biological neuroscience and AI, which serves as the theoretical foundation for Zyphra’s model architectures.

Zyphra has achieved significant technical milestones through a deep integration with the AMD hardware ecosystem, as detailed in the company’s research documentation.

Financial data from PitchBook indicates that Zyphra is currently a venture-backed company that attained “Unicorn” status in June 2025 following a $110 million Series A funding round. According to PitchBook and company press releases, Zyphra is supported by a group of strategic investors including Advanced Micro Devices (AMD), IBM, Bison Ventures, and BC VC. With a team of approximately 31 employees as of 2026, the company continues to expand its footprint through the Zyphra Inference Cloud and Maia, an intelligent assistant platform designed to bring advanced search and productivity tools to enterprise teams.

Community reactions and industry context

The announcement has resonated strongly within the AI community, garnering nearly 1 million views on X/Twitter within 24 hours. The excitement largely centers on two factors: the viability of the AMD stack and the efficiency of the reasoning “cascade.”

Technologists have noted that Zyphra’s post-training process—a 4-stage RL cascade—is unusually disciplined. Most labs use a single round of RL, but Zyphra’s pipeline includes a “reasoning warmup” followed by a curriculum of 400 adaptive puzzle-like environments (RLVE-Gym) before finally moving to behavioral polishing.

One of the most praised “under-the-hood” details is Router Replay. In MoE models, training can become unstable if the “trainer” engine and the “inference” engine make slightly different decisions about which expert to use for a token due to floating-point noise. Zyphra’s system records the exact expert choices made during generation and forces the trainer to use them, effectively “pinning” the computation path and ensuring higher learning stability.

As the industry faces a potential plateau in the benefits of simply adding more parameters, ZAYA1-8B provides a compelling counter-narrative: that the next frontier of AI isn’t just about bigger clusters, but about smarter “thinking” algorithms that can do more with less.

Elon Musk’s legal effort to dismantle OpenAI may hinge on how its for-profit subsidiary enhances or detracts from the frontier lab’s founding mission of ensuring that humanity benefits from artificial general intelligence.

On Thursday, a federal court in Oakland heard a former employee and board member say the company’s efforts to push AI products into the marketplace compromised its commitment to AI safety.

Rosie Campbell joined the company’s AGI readiness team in 2021, and left OpenAI in 2024 after her team was disbanded. Another safety-focused team, the Super Alignment team, was shut down in the same time period.

“When I joined it was very research-focused and common for people to talk about AGI and safety issues,” she testified. “Over time it became more like a product-focused organization.”

Under cross-examination, Campbell acknowledged that significant funding was likely necessary for the lab’s goal of building AGI, but said creating a super-intelligent computer model without the right safety measures in place wouldn’t fit with the mission of the organization she originally joined.

Campbell pointed to an incident where Microsoft deployed a version of the company’s GPT-4 model in India through its Bing search engine before the model had been evaluated by the company’s Deployment Safety Board (DSB). The model itself did not present a huge risk, she said, but the company needed “to set strong precedents as the technology gets more powerful. We want to have good safety processes in place we know are being followed reliably.”

OpenAI’s attorneys also had Campbell admit that in her “speculative opinion,” OpenAI’s safety approach is superior to that at xAI, the AI company that Musk founded that was acquired by SpaceX earlier this year.

OpenAI it releases evaluations of its models and shares a safety framework publicly, but the company declined to comment on its current approach to AGI alignment. Dylan Scandinaro, its current head of Preparedness, was hired from Anthropic in February. Altman said the hire would let him “sleep better tonight.”

The deployment of GPT-4 in India, however, was one of the red flags that led OpenAI’s non-profit board to briefly fire CEO Sam Altman in 2023. That incident took place after employees including then-chief scientist Ilya Sutskever and then-CTO Mira Murati complained about Altman’s conflict-averse mangement style. Tasha McCauley, a member of the board at the time, testified about concerns that Altman was not forthcoming enough with the board for its unusual structure to function.

McCauley also discussed a widely-reported pattern of Altman misleading the board. Notably, Altman lied to another board member about McCauley’s intention to remove Helen Toner, a third board member who published a white paper that included some implied criticism of OpenAI’s safety policy. Altman also failed to inform the board about the decision to launch ChatGPT publicly, and members were concerned about his lack of disclosure of potential conflicts of interest.

“We are a non-profit board and our mandate was to be able to oversee the for-profit underneath us,” McCauley told the court. “Our primary way to do that was being called into question. We did not have a high degree of confidence at all to trust that the information being conveyed to us allowed us to make decisions in an informed way.”

However, the decision to boot Altman came at the same time as a tender offer to the company’s employees. McCauley said that when OpenAI’s staff started to side with Altman and Microsoft worked to restore the status quo, the board ultimately reversed course, with the members opposed to Altman stepping down.

The apparent failure of the non-profit board to influence the for-profit organization goes directly to Musk’s case that the transformation of OpenAI from research organization into one of the largest private companies in the world broke the implicit agreement of the organization’s founders.

David Schizer, a former Dean of Columbia Law School who is being paid by Musk’s team to act as an expert witness, echoed McCauley’s concerns.

“OpenAI has emphasized that a key part of its mission is safety and they are going to prioritze safety over profits,” Schizer said. “Part of that is taking safety rules seriously, if something needs to be subject to safety review, it needs to happen. What matters is the process issue.”

With AI already deeply embedded in for-profit companies, the issue goes far beyond a single lab. McCauley said the failures of internal governance at OpenAI should be a reason to embrace stronger government regulation of advanced AI—”[if] it all comes down to one CEO making those decisions, and we have the public good at stake, that’s very suboptimal.”

Google revealed its new Fitbit Air today, and it’s the smallest tracker yet, designed exclusively for people who want some consistent health information without the hassle of a large, flashy device getting in the way. At only a little more than five grams on its own and around 12 grams with a band, this gadget glides onto your wrist and stays there for days on end with no problems. The engineers designed it to resemble a smooth pebble that nestles against the skin, and they managed to incorporate at least 35% recycled material by weight.

You have three band options to choose from right away: a recycled performance loop that has been a game changer so far, a sweatproof silicone active band for more physically demanding activities, and a sleeker elevated modern style that looks almost like real jewelry rather than a wearable. The first thing you’ll notice about this Fitbit Air is how low-key it is; there are no flashing lights, no obnoxious vibrations to disturb your day, and no large display to continuously check the time or stare at notifications.

Google Fitbit Air – Screenless Activity Tracker with Fitness, Heart Rate, and Sleep Tracking…

• Google Fitbit Air is the unbelievably comfortable, exceptionally smart way to transform your health[1]; and Google Health brings together effortless…
• Unlock more with Google Health Premium: With a premium membership, get personalized coaching that’s built with Gemini and adapts to your life…
• Comfortable fit – One Size Tracker (130-210 mm): The lightweight, micro-adjustable fit sits comfortably and quietly, so you can wear Google Fitbit Air…

With no display, you won’t have to worry about your phone ringing every two minutes. Instead, it silently gathers all of this data 24 hours a day, seven days a week and transmits it directly to your phone. Heart rate is continuously recorded, as are irregular heart rhythm alarms, blood oxygen levels, resting heart rate and heart rate variability, as well as the standard sleep stage tracking and workout recognition, all without the need for any fussing. Over time, the system begins to establish these patterns, providing you with a more full picture of your recovery and daily mobility.

Power-wise, you can expect it to last a full week on a single charge under normal use, with only a 5-minute top-up required to acquire another day’s worth when you run out of time. Water resistance is also adequate, up to 50 meters, so it can withstand showers, swims, and sweaty workouts without issue. The magnetic charger is easy to attach and reversible, so you won’t have to look for it in the dark. All of these small elements add up to a device that simply sits on your wrist without drawing attention to itself.

Data begins to flow into the all-new Google Health app, which now serves as a central hub for all of your Fitbit Air data, but it does more than that! The software collects a variety of different data, such as your phone’s health records or connected medical records, and then displays some excellent daily summaries, a weekly or monthly trend view, and some good ideas for how to improve. There’s also a coach tool built in, powered by Google AI, that generates entirely personalized training regimens depending on your recent sleep quality, heart rate recovery, or even if your schedule has changed. You can ask the coach what’s going on with this fatigue you keep feeling after a trip or how you should alter your exercises when you’re nursing an injury, and it will spew out all sorts of specific steps based on all of the data from this Fitbit Air

Pricing maintains the low entry point, at $99.99, which includes three months of Google Health Premium for free. There’s also a Stephen Curry-themed variant that costs $130 and includes a distinctive band design. Orders begin immediately, while the devices are slated to arrive in the United States and a few other locations beginning May 26th. The standard colors are obsidian, fog, berry, and lavender, but availability may vary by region. It’s all good for the latest and greatest Android phones starting with version 11, as well as the latest iPhones running iOS 16.4 or later. Setting it up is simple: all you need is a Google account and the Health app, and the tracker will quickly pair with your device over Bluetooth.

from the it’s-a-true-mystery dept

John Roberts has spent years whining about how totally unfair it is that people claim he and his colleagues rule based on partisan leanings. He did it in 2014. He did it in 2017. He did it in 2019. Hell, he did it a couple months ago too. So it’s little surprise that he’s out there whining about people calling the Court partisan yet again.

Speaking at a conference for lawyers and judges in Hershey, Roberts said the Supreme Court is required to make decisions that are not popular and bemoaned that there is not a better understanding among the public of how the court operates.

“I think at a very basic level, people think we’re making policy decisions, [that] we’re saying we think this is what things should be as opposed to this is what the law provides,” Roberts said. “I think they view us as truly political actors, which I don’t think is an accurate understanding of what we do. I would say that’s the main difficulty.”

While he conceded that people have a right to criticize the court and its decisions, he added that there is a tendency to focus too much on politics.

“We’re not simply part of the political process, and there’s a reason for that, and I’m not sure people grasp that as much as is appropriate,” Roberts said.

Advertisement

The timing here is something else — a week after an obviously partisan ruling in Callais, which stripped away Section 2 of the Voting Rights Act. Notably, Roberts himself had pointed to Section 2’s existence back in 2013 as the reason that they could kill off Section 4 of the Voting Rights Act (which required a pre-review of voting maps for racial bias). And now he helped kill Section 2.

If it were just about making decisions that are “not popular,” then… why are nearly all of his “unpopular” decisions quite clearly in support of one party’s goals and ideology? Any look at the details shows why people conclude that Roberts has a partisan bent to his rulings:

• In the 15 precedent-overturning cases with partisan implications, in other words, Justice Roberts voted for a conservative outcome 14 times (93%).
• Chief Justice Roberts is one of only two justices since 1946 to support 100% of decisions overturning precedent that led to conservative outcomes.
• Roberts’s record in precedent-overturning cases is the second-most conservative among 37 justices who have ruled in at least 5 precedent-overturning cases since 1946. With 84% conservative votes in precedent-overturning cases, Roberts only trails Justice Alito’s 88%.

Gee. I wonder why people think the Court is partisan, chief?

And, on Monday (as we pointed out) Roberts joined Alito and the conservatives on the bench to break standard practice and precedent, supporting Louisiana ripping up its election maps to favor more Republican seats — even as voting had already started — even though, just months ago, he and the conservatives had said that Texas’ map (deemed unconstitutionally based on race by a Trump-appointed judge) couldn’t be torn up because it was “too close” to an election and voters needed “certainty.” There is literally no explanation for December being too close to change the maps while May somehow required rushing a map change… in the same election… other than the partisan leaning of those two decisions.

Indeed, as Liz Dye points out, we have decades of the Supreme Court doing exactly this: it allows for election map changes when it will help Republicans, but says “no can do, too close to an election” whenever it’s expected to help Democrats:

The Court’s conservatives routinely scold lower court judges for changing voting rules too close to an election. This violates the Purcell principle, named for a 2006 case in which the Court rebuked the 9th Circuit for blocking Arizona’s voter ID law too close to an election and causing voter “confusion.” For 20 years, the Supreme Court’s conservatives have selectively invoked Purcell to allow elections to proceed using maps that courts have already deemed to be unlawful.

In 2022, after lower courts struck down Alabama’s electoral map for violating Section 2 of the Voting Rights Act and disenfranchising Black voters, the Supreme Court intervened to allow Alabama to use the unconstitutional map anyway in the midterms. In 2023, the Court agreed that the maps were illegal under the VRA — but only after they’d let Alabama Republicans use them to take back the House.

Just five months ago, the Court cited Purcell when it rebuked a federal district court for “improperly inserting itself into an active primary campaign” by blocking Texas’s unconstitutionally racial gerrymander.

But given the chance to insert themselves into an acting primary campaign, they regularly jump in with both feet. And in fact they’re equally happy to stomp into the primary itself.

So, chief, if you want people to stop thinking the Court is partisan, maybe stop making such obviously partisan decisions.

Oh, and also maybe talk to your colleagues. After all, at the very moment you were whining about people thinking the court was partisan, your colleague Justice Neil Gorsuch was appearing on a famously rightwing podcast to talk about why “young conservatives must have courage to stand by their beliefs.” Sounds kinda partisan.

And just a few weeks ago, your colleague Justice Clarence Thomas gave a speech arguing that progressives were an existential threat to America.

Gosh. Why would the public think some of you are partisan. I wonder!

And, let’s not forget that Thomas’s wife was supportive of the attempt to steal an election from the rightfully elected Joe Biden in support of the failed Republican campaign of Donald Trump. And then there’s Justice Alito’s wife who, somewhat infamously, flew political flags outside their home, including one in support of the January 6th insurrection.

A real mystery, truly. Who could possibly think that there might be partisan bias? How unfair.

But you keep saying how unfair this is. Year after year, conference after conference, the same complaint: people just don’t understand us.

At some point, Chief Justice, the more productive question isn’t why the public doesn’t grasp your supposed non-partisanship. It’s why — after decades of rulings that break almost exclusively in one direction, colleagues who deliver speeches about the courage of young conservatives, and the existential threat of progressivism, and spouses flying insurrection flags — you’re still surprised that they don’t.

Maybe the problem isn’t the public’s understanding. Maybe it’s the Court’s behavior.

Filed Under: clarence thomas, john roberts, partisanship, samuel alito, supreme court

The Global Positioning System (GPS) was developed by the United States military in the 1970s, but it wasn’t long before civilians all over the planet started using it. By the early 2000s the technology was popping up in consumer devices such as mobile phones, and since then its become absolutely integral to our modern way of life.

But although support for GPS in our gadgets is nearly ubiquitous, it’s not the only option when it comes to figuring out where you are on the globe. As you might imagine, not everyone was thrilled with building their infrastructure around one of Uncle Sam’s pet projects, and so today there are several homegrown regional and global satellite navigation systems in operation.

As a follow-up to our recent dive into the ongoing GPS upgrades, let’s take a look at some of the other satellite positioning systems and who operates them.

GLONASS (Russia)

Given the tensions of the Cold War, it will probably come as little surprise to learn that the Soviet Union introduced their own satellite-based navigation system to compete with GPS. Development of the Global Navigation Satellite System (GLONASS) started a few years later than its Western counterpart, with the first satellites not reaching orbit until 1982, officially making it the second Global Navigation Satellite Systems (GNSS) ever developed.

GLONASS satellites orbit at a slightly lower altitude than GPS, 19,100 kilometers (11,900 miles) compared to 20,200 km (12,600 mi) of the American system, and at a greater inclination. This makes reception better at higher latitudes, which makes sense given the desired coverage area.

As designed the capabilities and overall accuracy of GLONASS were very similar to GPS, but the early satellites had a short operational lifespan of just three years. For global coverage GLONASS required 24 satellites in orbit, and maintaining coverage over Russia required 18. But after the fall of the USSR, launches of new satellites were put on pause and the constellation started suffering losses. By 2001, there were just seven operational GLONASS satellites.

President Vladimir Putin made the restoration of GLONASS a key priority in his administration, leading to resumed launches and development of the second and third generation satellites. Within a few years, commercial interest in GLONASS started to pick up, and the network regained global coverage in 2011. While the constellation has experienced a few setbacks over the last several years, spare and replacement satellites have been launched regularly, with the most recent entering orbit in September of 2025.

BeiDou (China)

Unlike the American and Russian systems, the first iteration of BeiDou was of a much smaller scale. Rather than a global system, the goal was to provide regional coverage for China and the surrounding countries with just four satellites placed in a geostationary orbit at an altitude of approximately 35,786 km (22,236 mi). From an observer in China, the satellites would appear to be motionless in the sky, ensuring reception anywhere in the country. Known retroactively as BeiDou-1, the system was operational from 2003 to 2012.

That year it was replaced with the far more ambitious BeiDou-2. The design called for a constellation of satellites in various orbits: 5 geostationary to provide backwards compatibility with BeiDou-1, 27 in medium Earth orbit similar to GPS/GLONASS, and 3 in an inclined geosynchronous orbit. The latter meaning that from the perspective of Earth, the satellite would appear to loiter overhead rather than remain in a fixed position.

BeiDou-1 was largely a research project and saw little use outside of the Chinese government. Conversely BeiDou-2 was designed for both government and civilian use from the start, with two distinct levels of service — civilian users could plot their position within a radius of 10 meters (32 feet), while the military reportedly enjoyed an accuracy of 10 cm (4 inches).

The coverage area of BeiDou-2 was expanded considerably to the south to include include Indonesia and Australia, but it still didn’t provide global service. Commercial use of the network started to pick up at this point, and by 2014 smartphones from Sony, Samsung, and Xiaomi included support for it.

It wasn’t until the introduction of BeiDou-3 in 2015 that the system could boast global coverage, with the system reaching full operational status in June of 2020.

Galileo (European Union)

While civilian use of GPS, GLONASS, and BeiDou was always part of the plan, all three systems were ultimately designed  as tools of their respective governments. Conversely, when the European Union set out to develop Galileo in the early 2000s, the goal was to create a satellite navigation system operated by private companies and aimed at civilian users.

That first part of the plan fell apart fairly quickly, and by 2006 Galileo was nationalized and the European Space Agency was entrusted with its development and operation. The first operational satellite was put into orbit in October 2011, and limited functionality was available to the public by 2016. While Galileo was designed for civilian use, it does offer a High Accuracy Service (HAS) with an accuracy of 20 cm (8 inches) that was initially intended to be accessible only by paying customers. But eventually it was decided to make HAS available to compatible receivers free of charge. When combined with its interoperability with GPS and GLONASS, Galileo offers exceptional accuracy.

Galileo reached full operational status in 2024 with a constellation of 24 satellites. Starting in 2027, these will be joined by a dozen upgraded Galileo Second Generation (G2) satellites that feature more electric propulsion for more efficient orbital maneuvers, improved antennas, and inter-satellite data links.

QZSS (Japan)

Development of the Quasi-Zenith Satellite System (QZSS) started in 2002, with the goal of offering high-accuracy position services to users in and around Japan. But rather than operating independently, QZSS was designed to augment GPS with five additional satellites.

Two of the satellites are in a geostationary orbit similar to those used in China’s BeiDou-1 system, while the other three are in a geosynchronous orbit like those introduced with BeiDou-2. These orbits are intended to keep at least one satellite directly over Japan at all times to improve reception in urban areas. The system became fully operational in 2018.

In the near future, Japan plans on adding three more satellites to the QZSS constellation. This would give the system enough regional coverage to operate independently of GPS if necessary.

NavIC (India)

Navigation with Indian Constellation (NavIC), previously known as Indian Regional Navigation Satellite System (IRNSS), is an independent regional navigation system that covers India and the surrounding area using seven satellites.

Development of NavIC started in 2006, and the first satellite was launched in 2013. Like QZSS, the constellation is made up of satellites in both geostationary and geosynchronous orbits. Two levels of service are offered: the Standard Positioning Service for civilian use that offers an accuracy of 3 m (9.8 feet), and an encrypted Restricted Service intended for military and government applications that’s accurate to 2 m (6.7 ft)

One of the goals of NavIC was not only to launch and operate the system from within India, but to produce as much of the hardware domestically as possible. This includes the atomic clocks and microprocessors aboard each satellite as well as the receiver chips used in client devices. While India wanted to maintain ultimate control over NavIC for political reasons, it’s not an isolationist system — it is designed to be interoperable with other GNSS.

That last point is particularly important right now, as only three NavIC satellites are currently transmitting navigational data due to hardware issues. Those three satellites alone aren’t enough to plot an accurate position, so to compute their location receivers must pull in data from other systems such as GPS.

Better Together

Although having so many active satellite navigation systems may seem redundant, the fact that they all offer at least some level of interoperability with each other means that everyone with a multi-system receiver can benefit. Instead of being limited to the constellation of just one service, this cross compatibility lets a device pull in data from whatever satellites are overhead at the time.

Granted how much of an improvement this results in will be highly dependent on where you’re located on the globe, but no matter what, its always going to be better than being limited to just one system.

Singapore’s low TFR is more than just a financial problem

Singapore’s total fertility rate (TFR) hit an all-time low of 0.87 in 2025. To put that in perspective, a country needs a TFR of at least 2.1 to maintain its population without immigration. Singapore is now well below that threshold—and the decline shows little sign of slowing.

Deputy Prime Minister Gan Kim Yong called it an existential issue, warning that “over time, it will be practically impossible to reverse the trend because we will have fewer and fewer women who can bear children.” 

Demographers project that by 2050, seniors aged 65 and above could make up half of Singapore’s population, while the working-age population may shrink by 16%. This is closely tied to falling birth rates: when fewer children are born over time, there are eventually fewer people entering the workforce to replace retiring generations.

As the population ages and the pool of working adults shrinks, a smaller base of taxpayers must support a growing elderly population, placing increasing strain on public finances and the broader economy.

Singapore has been aware of this problem for decades.

It has spent billions trying to fix it—annual spending on marriage and parenthood support alone quintupled to S$2.5 billion between 2001 and 2017, and has only grown since. But the results have been largely limited. Every prime minister since Lee Kuan Yew has tried to move the needle, but the TFR has only continued to decline from 1.82 in 1980 to 0.87 in 2025.

This raises an uncomfortable question: monetary aid is certainly helpful, but is the government actually addressing the right problem?

What the government has done

The Singapore government’s response to falling birth rates has been, by any measure, substantial, and its Marriage & Parenthood Package under the Made for Families Initiative is widely recognised by the United Nations as one of the “most comprehensive pronatalist programmes in Asia,” covering everything from cash transfers to housing, parental leave, and childcare.

For every newborn, parents receive a Baby Bonus Cash Gift of S$11,000 for the first and second child, a S$5,000 First Step Grant in a government co-savings account, a S$5,000 MediSave Grant for healthcare, and a S$5,000 Parenthood Tax Rebate. In total, this amounts to a baseline of around S$26,000 in direct benefits for a first child alone, before factoring in additional childcare subsidies, housing grants, and education-related support.

On parental leave, working parents now have a combined total of 30 weeks of paid leave, comprising 16 weeks of maternity leave, 4 mandatory weeks of paternity leave, and a new 10-week Shared Parental Leave pool.

Beyond this, the government has also expanded childcare infrastructure, doubling full-day preschools from around 100,000 in 2013 to over 200,000, and progressively reduced fee caps at government-supported centres. 

On education, Singapore Citizens pay no monthly school fees at the primary level, and just S$5 a month at secondary school. According to SingStat data sourced from the Ministry of Education, the government spends approximately S$14,541 per primary school student and around S$18,619 per secondary school student annually. Fees only begin to rise meaningfully at the tertiary level, where polytechnic and university costs step up significantly.

Meanwhile, priority housing schemes give married couples with children earlier access to public housing, and the Tripartite Guidelines on Flexible Work Arrangement Requests took effect in Dec 2024, requiring employers to properly consider caregivers’ requests for flexible arrangements.

On paper, this is an extraordinarily generous package by the Singapore government. And yet, the TFR keeps falling.

Singapore is not alone in trying all these monetary measures. Taiwan, South Korea, and Japan have all tried different versions of the same playbook—cash transfers, longer parental leave, more childcare—with equally disappointing results. 

Financial incentives alone cannot fix what is fundamentally not a financial problem, despite the high cost of living being a key consideration in having a child in Singapore.

The actual problem

There is a fundamental flaw in Singapore’s approach: every single one of these policies is targeted at people who have already decided to have children.

The Baby Bonus helps you once you’re pregnant. Parental leave supports you once a child is born. Childcare subsidies reduce the cost of raising a child you’ve already had. Even the housing priority scheme assumes you’ve already committed to starting a family.

Essentially, the government is optimising the experience of parenthood for those who have already chosen it, but doing almost nothing to address the far larger and more critical group: people who have decided they don’t want children at all.

The numbers bear this out. Singapore’s TFR problem is not primarily that parents are having fewer second or third children (though that matters, too). It’s that a growing proportion of Singaporeans are simply not becoming parents in the first place. 

The proportion of married women without children has more than doubled from 6.7% in 2004 to 14.4% in 2024. Total marriages registered fell 7% in 2024, driven by fewer unions among those aged 25–34—the prime family-formation years. At the same time, the median age at first marriage has risen steadily, from 28.2 years for brides in 2014 to 29.6 years in 2024.

Taken together, these shifts point to a delayed and shrinking window for family formation, which helps explain the long-term decline in birth rates. But beyond delays alone, a growing number of young Singaporeans are not just postponing parenthood—they are choosing to forgo it entirely.

The government has made some cultural gestures, especially through the Families for Life community programmes, to promote a “family-friendly culture ” through bonding activities and supportive resources.

However, these are surface-level nudges rather than structural interventions. They assume the desire for children already exists and just needs encouragement. In doing that, they fail to grapple with the far harder question: what happens when that desire has been eroded entirely?

Moreover, the political discourse hasn’t helped. Conversations about marriage and parenthood in Singapore have historically been framed in transactional terms—what you get, what it costs, what the government will subsidise. When the narrative around having children is built on spreadsheets rather than meaning, it’s unsurprising that people run the numbers and opt out.

Why Singaporeans really don’t want children

The real reasons more Singaporeans are choosing not to have children go beyond monetary concerns. They are psychological, philosophical, and cultural, and cannot be solved by the mere gift of a larger Baby Bonus. 

1. Philosophical anxiety about a pessimistic future & generational trauma

A growing number of young Singaporeans carry a quiet but real fear: “What kind of life am I bringing a child into?”

This is largely not irrational, but a genuine reckoning with the Singapore they have inherited—high costs of living, relentless pressure to perform, a society where the path from birth to CPF feels pre-scripted and exhausting from the societal rat race.

“Do I really want to give life to a generation that will have to deal with what I am dealing with now that may only get possibly worse?” one netizen commented on an Instagram post by Rice Media discussing Singapore’s fertility rate. While not representative of everyone, it captures a sentiment that surfaces frequently, even if not always openly expressed.

There is also a growing awareness of generational trauma: the patterns of emotional unavailability, perfectionism, and anxiety that parents pass down to children (evident from the countless extracurricular classes, tuition and enrichment lessons children here are sent to)—and a conscious choice by some not to continue that cycle.

No financial incentive can speak to existential doubt this deeply.

2. Economic rationality & the real cost of tradeoffs

Yes, raising a child is expensive. But the bigger issue isn’t the price tag—it’s everything else you’re trading away to have one.

A generation of Singaporeans, more educated and economically empowered than any before them, now has more compelling alternatives to parenthood than ever: travel, career progression, financial independence, and personal freedom.

A child is not just a financial expense—it is a permanent reallocation of one’s time, identity, and autonomy.

When you have worked hard to build a life you enjoy, and you can see clearly what a child would cost you in terms of sleep, career, relationships, and spontaneity, the calculus looks very different from what it did for previous generations, for whom alternatives were fewer and social expectations of a nuclear family were much more strongly enforced.

3. Work culture & the myth of balance

Singapore’s work culture is often described as intense and demanding.

Several reports have found that Singapore is the most overworked country in the Asia Pacific Region. In fact, it has the longest working hours per week at 45, followed by China at 42.

In the Rice Media Instagram post, another netizen even noted that they only get about an hour a day with their children due to the demands of work.

Even where parental leave exists on paper, its use is often shaped by workplace culture. Research by the Ministry of Social and Family Development found that a key reason some fathers who wanted to take paternity leave did not do so was concern over career repercussions, workload implications, or the burden placed on colleagues during their absence.

Beyond the workplace, the “second shift” of parenthood is just as demanding.

The mental load of raising a child in Singapore—navigating an exam-driven education system, managing enrichment schedules, and constantly worrying about a child’s future in an intensely competitive environment—effectively becomes a full-time job layered on top of an already exhausting full-time job.

For dual-income couples already stretched thin, this is not a problem that can be solved simply by adding more childcare slots or financial incentives. It goes deeper: a question of whether the structure of work and life itself leaves enough space for parenthood to feel sustainable—or even desirable.

4. The era of non-commitment

Relationships are the foundation of family formation, and the landscape in Singapore has shifted dramatically.

The rise of situationships, dating app culture, and a general wariness around long-term commitment among younger Singaporeans means that the pathway to marriage—historically the precursor to children in Singapore’s social context—is less travelled and more delayed than ever.

This is partly economic (settling down is expensive), partly cultural (independence is more valued), and partly a reflection of the same risk-aversion that shows up elsewhere. 

Committing to a partner, like committing to a child, is a bet on an uncertain future in an environment that feels precarious.

Resetting the narrative surrounding parenthood

The question we should be asking is not “how do we make having children cheaper?” but rather: why did people once imagine having children, and why have they stopped?

Answering that requires looking at the lived experience of growing up and working in Singapore. The education system that measures worth in grades and portfolios from age seven. The social media landscape that makes childlessness look like freedom and parenthood look like sacrifice. The working adult’s calendar, where evenings and weekends are either recovery time or resume-building. 

The question of whether there are spaces in Singapore to simply enjoy being alive, and whether those spaces are accessible without being expensive.

If children growing up in Singapore feel that their childhood was a series of performance reviews, they will not look back on it with warmth. And people do not recreate experiences they do not value.

The image of parenthood changes when the image of childhood changes. That means engaging seriously with education reform—moving away from a system built on academic anxiety toward one that allows children to actually enjoy their youth. 

It means addressing work culture honestly, not just at the policy level but at the level of employer norms and social expectations. It means making space for young people to form genuine, committed relationships—and examining what cultural and structural forces are working against that.

None of this is easy, and it will take more than budget reforms to do the trick.

• Read other articles we’ve written on Singaporean businesses here.