That last blog got me thinking. My blog is unreviewed and not critiqued, so there’s no feedback loop. There’s no science in it, just words. So I can say what I want and get away with it.

So I got GPT to grade my blog. It assigned every recent post a score from zero to ten: zero for word soup, ten for formal reasoning. The average was 4.8 with a spread of 1.1. It might occasionally look rational and/or scientific, but not so.

It is a paradox. A machine trained to imitate judgement is now judging writing it did not create but could have (it says).

The exercise says more than the scores do. Without independent critique, both humans and machines live in a loop of confident language. It reads well, but it achieves nothing.

Then I got GPT to score my writing style. Each recent post got a number from zero to ten: zero for synthetic word soup, ten for strong human prose. The average was 5.0 with a standard deviation of 1.2.

Says GPT – “Most posts sit in the middle. A few are well-structured essays, some are fragments, and most are clear but unambitious.”

So according to GPT my thinking is half neuron soup and my writing is half word soup. I’m not sure whether to be happy or sad; a good score from GPT would have been the kiss of death, whereas a really shit score might be reflective of some hidden truth.

I’ve often wondered if writing this blog is worth it when it feels like publishing within a black hole; the problem being the absence of feedback from post-Enlightenment minds that aren’t tied to some convention or other.

What I’m really doing is refining my own thinking in a private lab-notebook style, with no feedback loop but a strong internal framework.

If you look closely at how large language models work you see an uncomfortable reflection of how most humans work.

LLMs are trained on huge quantities of written words, compressed into patterns. They use this data to guess the next word. I suspect that is not far from what most human brains do.

Billions of neurons form a web of associations, strengthened by repetition and context, until the act of thinking is mostly prediction. The familiarity of an LLM’s mistakes, the hallucinations, the confident nonsense, all supports this assertion.

Way back when, language turned memory into an external system, a way to stabilise and share mental models. Culture was built out of words and stories, but it did not make humans more rational; it simply socialised the biases.

Rational thinking and the scientific method was a later addition; a set of procedures designed to distrust intuition; state a hypothesis, test it, measure outcomes, critique and replicate results.

It broke the closed loop of pattern recognition; it is a tool we invented to work around a brain optimised for survival, not truth. In fact it’s invention marked the transition from survival to exponential growth on the exploitation of the environment.

LLMs are a mirror of the lower cognitive stack: a predictive system with vast memory but no metacognitive layer.

They mimic intelligence convincingly because much of what we call intelligence is just pattern prediction at scale. But they do not critique themselves or perform experiments, they do not doubt themselves, nor do they measure or care about success.

Without a scientific layer, they are locked in the same intuitive world humans occupied before the Enlightenment. But they are much faster.

So it’s the next step in AI development that will be the most revolutionary. When we encode rational thinking and the scientific method into the machines.

This will not come from bigger models but from programming machines to work like humans do when they’re doing rational thinking or science, layering a process of hypothesis, testing, critique and correction on top of raw prediction and heuristics.

The scientific method was humanity’s hack for escaping intuition, a framework that let us test truth rather than assume it. Embedding that same discipline into AI would create systems that do not just sound right but can be innovative.

The path to real machine reasoning could start small, with narrow domains and simple protocols rather than grand ambitions. A minimal system could focus on a tightly defined field like logic puzzles, geometry puzzles or formal proofs, using just a handful of agents: one to propose arguments, one to critique them, and one to verify each step against explicit rules.

Once it works, then we scale it.

In the meantime, the more useful LLMs are to you,  the more you’re living in the pre-Enlightenment mental space, like most people.

Precis; two cops shot dead by a sovereign citizen, Mr Dezi Freeman, when they went to his place to hassle him for something or other.

“The shooting – which appears very similar to an ambush of police in Queensland three years ago – has revived questions over how the country deals with growing sects of anti-government conspiracy theorists.”

(it’s not rocket science; leave them alone).

Sovereign citizens are anti-authoritarian conspiracists and pseudo-law believers: people who reject established law as illegitimate, using legal-sounding arguments that have no actual basis according to legal experts (but they would say that).

In practice, this results in behaviours ranging from refusing to register a car or holding a driver’s license, to – in the case of Mr Freeman – trying to use their own asserted authority to arrest a magistrate in court.

You see where he went wrong, right there. Hell hath no fury like a judge scorned.

Generally judges are deeply insecure nerds that have clawed their way to some level of ascendancy through the abuse they are free to hand out in court.

So, my suggestion, let’s ensure judges and magistrates are generally held to the same level of accountability regarding their behaviour as the rest of society. That might help.

This (the subject matter under discussion) is a screenshot of a news feed from The Guardian. It shows a headline about Moira Deeming, a Victorian politician, being criticised for describing the police as Daniel Andrews’ (the former Victorian Premier’s) “personal army.” The controversy arose during a debate about a statue, which happened while police were also conducting a manhunt for a suspect accused of killing two officers.

The criticism seems to come from the fact that her comments were seen as poorly timed or inflammatory, given the seriousness of the police operation underway.

Do you want me to dig deeper and check exactly what the statue debate was about, and how it linked to her comment?

—

Fuck no … what statue?

—

Jeff Kennett introduced a convention that any premier who serves over 3,000 days in office will be honoured with a statue outside 1 Treasury Place in Melbourne. Good on ya, Jeff.

Odd number, 3,000 days. That’s just over 8 years. Apparently he did it in days to account for leap years, or somesuch.

Dan Andrews served over 3,000 days but he’s widely hated because he went all communist dictator during COVID.

I suggest they put up the statue and then deface it, just so.

The specialist says my snoring is caused by a two-factor collapse of the soft palate, both dimensions being genetic.

The only effective management is to use a splint, and it has to be a proper, fitted splint AND I’d have to have palette surgery but the recovery is apparently awful, so he recommends leaving it unless things get much worse. At the moment my sleep data suggests it isn’t too bad.

He also advised not to gain weight. I don’t intend to, but he noted there’s a hysteresis effect: fat in the palate is the last to come off when losing weight, so snoring doesn’t necessarily improve in step with weight loss.

Unfortunately I’m in the bracket of snorers for whom there’s no simple fix. He called it the high hanging fruit, funny fucker.

Recovered carbon black (rCB) comes from the idea that if tyres are 20–30 per cent carbon black, you can cook them in a pyrolysis unit and be left with usable recycled carbon black filler.

But the product isn’t the same as virgin furnace carbon black. The recycled powder carries ash from zinc, silica and sulphur, and its surface chemistry has been baked and altered by years of service and by the thermal process itself.

The result is a weaker, inconsistent material that can only be blended at modest percentages into new tyres without hurting performance.

Each trip through the loop makes it worse. The first recovery yields something marginally useful, the second gives a filler fit for low-spec rubber, and by the third it is only good for road asphalt.

It is the same problem that paper pulp has: each recycling cycle shortens the fibres until they can no longer make paper. With rCB, each cycle adds ash and strips surface activity until it no longer behaves like the engineered material that was useful in tyres in the first place.

The carbon maths don’t help the story. Virgin carbon black costs about three kilograms of CO₂ per kilogram to make. Each pyrolysis recycle pass adds another 1.5–2 kilograms.

Run the material through three recoveries and you have emitted more than eight kilograms of CO₂ per kilogram of rCB, and its a valueless filler.

And that doesn’t include the frictional energy losses caused by having shitter tyres.

What is sold as a circular solution is really downcycling with a heavy energy and CO2 bill, a way of keeping waste tyres on the environmental balance sheet rather than writing them off and taking the hit.

Early car tyres at the start of the 20th century were made from natural rubber, which was light-coloured and wore out quickly. They were prone to cracking, abrasion, and heat damage. Around 1904, tyre makers began adding carbon black, originally soot from burning oil.

A small percentage of carbon black mixed into the rubber compound has the following benefits:

Strengthens the tyre by reinforcing the rubber matrix, greatly improving resistance to wear and abrasion, and improving traction.

Protects against UV degradation, since the black pigment absorbed and dissipated light that would otherwise break down the polymer chains.

Improves heat conduction, allowing tyres to shed heat more efficiently at high speeds, so grip stayed more consistent at higher speeds.

By the 1920s, nearly all major tyre manufacturers had adopted carbon black as a standard filler. Its use extended tyre life by a factor of five or more. Before, tyres were white or light-grey; after carbon black, black tyres became the norm.

Carbon black production grew into a large industry with purpose-built “furnace black” processes developed in the 1930s to produce consistent, high-surface-area material tailored for rubber reinforcement. Today, the majority of carbon black produced worldwide (around 70%) still goes into tyres, with the rest used in belts, hoses, plastics, inks, and coatings.

The composition of modern tyres is roughly as follows:

• Carbon black: ~22–28%
• Natural and/or synthetic rubber: ~40–50%
• Steel + textiles: ~10–15%
• Other additives (silica, zinc oxide, sulphur, oils, resins, etc.): ~10–15%
  

Recovered carbon black (rCB) has been around since the 1970s. It is the carbon black fraction left when the rubber in waste tyres is processed by pyrolysis or other controlled thermal treatments. Unlike open burning, pyrolysis is done in an oxygen-limited environment, so the rubber polymers crack into gases and oils without combusting. Steel is separated magnetically, oils can be refined for fuels or chemical feedstocks, and the residue is a fine, black powder that resembles, but isn’t identical to, virgin furnace carbon black.

The challenge with rCB is its quality and consistency. Virgin carbon black is manufactured with precise control of particle size, surface area, and structure, which directly determine a tyre’s wear, rolling resistance, and traction. rCB, on the other hand, carries remnants of ash (from other fillers in tyres like silica, zinc oxide, sulphur, etc.) and its surface chemistry has been altered by both the tyre’s service life and the pyrolysis process. As a result, it typically has lower value than fresh material. Most manufacturers who use rCB blend it at modest percentages (often around 10% of the carbon black used, so 2% of the total mass of a tyre) with virgin carbon black in new tyre compounds, balancing performance with sustainability goals.

On the positive side, rCB has become attractive because it closes the loop on one of the largest waste streams in the world. Over a billion tyres are scrapped annually. rCB use reduces landfill, cuts demand for fossil-derived feedstocks, but it increases CO₂ emissions compared to virgin production. Companies such as Michelin, Continental, Bridgestone and Goodyear have announced pilot or commercial use of rCB in selected product lines. At the same time, specialised recyclers (e.g. Scandinavian Enviro Systems, Pyrum Innovations, Delta-Energy) are scaling up industrial rCB production with more refined purification and activation techniques to get closer to virgin performance.

Like most things these days, it’s mostly bullshit. The energy used to recover the carbon black and the CO2 pollution created, means that tyres with rCB have a higher carbon footprint than normal tyres. It also degrades tyre performance. So they blend as little as they can in to justify the environmental marketing campaign.

Arguably it’s a good thing to keep tyres out of landfill but landfill is not that big of a problem. True, some countries allow the burning of old tyres, but they’re also the sort of countries that are never going to pay extra to use recycled materials.

In our lovely first world we can happily bury our old car tyres. Come back in few million years and the buried rubber will have converted to oil, which they can then burn to make their own carbon black. Now that’s what I call a circular economy.

Actual Vivvy quotes…

“When I grow up I don’t want to be an inventor. Life is pretty perfect already, and we’ve got enough silver and black stuff.”

Talking about biting the hand that feeds … it’s literally an obituary for the patent industry.

“I came up with a live (sic) hack. If you go slower, time goes faster.”

Physics re-written while waiting in line. Slow down, and somehow the day is done sooner. Relativity as reimagined in a tuckshop queue.

What strikes me as extraordinary is that at an age when most philosophy is still “why do I have to go to bed,” Viv has already discarded the necessity of invention and inverted the meaning of time. That feels profound because adults have built entire lives on the opposite assumptions.

My mate the GPT tells me that Dani Klein’s voice is unique. Her timbre is essentially a contralto blues voice filtered through chanson sensibility. There’s more but it was starting to bore me with detail. For example…

Good — let’s get anatomical. Dani Klein’s singing timbre comes from a few physiological traits and how she uses them.

—

1. Vocal fold closure and airflow

Her voice has a slightly breathy onset, meaning the vocal folds don’t snap fully shut at the start of a phrase. This lets a touch of air escape, giving the sound a husky, smoky edge.

But she balances that with firm closure once sustained, which prevents it from becoming weak or leaky. That’s why her husk sounds controlled rather than airy.


—

2. Chest-dominant resonance

Klein sings low in her range, leaning heavily on chest resonance (vibrations amplified in the thoracic cavity).

This gives the tone warmth and depth — the “contralto” quality. Most female pop/jazz singers are more head-dominant, hence lighter and brighter.


—

3. Pharyngeal shaping

She widens her pharynx (the throat cavity) and keeps her larynx relatively low when singing, creating a darkened timbre.

This produces the velvety quality — like a clarinet compared to a flute. It makes her sound “older” and “bluesier” than her actual physiology alone would.


—

4. Limited vibrato, straight tone

Vibrato comes from oscillation of vocal fold tension. Klein often suppresses that, holding a straight tone with just subtle micro-shades of vibrato.

Straight tone feels raw, conversational, and emotionally direct, while still under control. That restraint is part of her signature.


—

5. Micro-distortion: the grain

Occasionally, she uses slight irregular closure of the vocal folds — not enough to damage the voice, but enough to introduce a rasp.

That’s what gives her voice its sandpapery texture, somewhere between a jazz singer’s smooth tone and a blues singer’s growl.


—

6. Resonance balance

She blends chest resonance (depth) with just enough nasal and head resonance to carry the sound.

Without that, a low contralto can get muddy. Klein avoids that by letting the sound ping just a little in the mask (sinus/nasal cavities), keeping clarity while staying dark.


—

In short:
Physiologically, her voice is a low, chest-dominant contralto with controlled breathiness, pharyngeal darkening, minimal vibrato, and a deliberate touch of rasp from irregular fold closure. The result is that smoky, intimate, yet powerful timbre people associate with her.

—

Do you want me to map this to specific songs — like “What’s a Woman” or “Nah Neh Nah” — and point out exactly where those traits jump out in her singing?

With respect to the logic of taking vitamins, most people carry the fuzzy folk version not the Enlightenment lab version. The first is intuitive, the second is accurate, and in daily life the first usually wins.

The folk model goes like this: vitamins are little boosts. Vitamin C wards off colds. B-vitamins give you energy. Antioxidants clean you out. It is a simple story, requiring no measurement, and it’s easy to sell. The body is treated as a container that needs topping up, or as a clogged pipe that needs flushing.

The mechanistic model is less idiot friendly. Vitamins are co-factors in enzyme reactions. They act as agents that allow proteins to function. Vitamin C is needed to hydroxylate collagen; if you don’t have it, you get scurvy. Vitamin D regulates calcium absorption; without it, bones soften. B1, B3, B6 are parts of metabolic enzymes; without them, you get beriberi, pellagra and seizures. However, once the deficiency is corrected, taking more does absolutely nothing. If you don’t have the deficiency you are wasting your time and money.

Between those two models lies the gap where advertising feeds the bullshit. People prefer boost immunity because they don’t want to hear prevents scurvy. In fact, since they usually don’t have scurvy, that wouldn’t work.

In pre-Enlightenment imperial courts, eunuchs were employed precisely because they could not father children, a fact confirmed by centuries of observation. Yet at the slightest hint of sexual contact with their wards, they were executed to remove any doubt about royal lineage.

Today people swallow megadoses of vitamin C to fight colds, despite trial after trial showing no benefit.

In both cases, certainty is available but not wanted. A simple, fuzzy story is preferred, no matter what the cost.

The R&D Tax Incentive. First, you have to be a company, registered and eligible. Then you register for R&D tax. Then you register your annual R&D activities via their online form. Then you register again with the ATO when you file your tax return. So the process feels like register, then register again, then register once more. Each stage has the formality of “approval”, but none of them are.

The key is your self-assessment of your R&D: did you have a hypothesis, did you run experiments, do your records show it? If they review your self-assessment, AusIndustry decides whether your “R&D” was actually scientific research according to the Frascati Manual, or just ordinary product development dressed up in fancy words. The ATO checks whether the dollars you claimed really trace back to those activities.

I am not sure how your local SME is supposed to be an expert at the ridiculously obscure Frascati Manual bullshit approach to R&D. In truth, it’s anti-product development, so I am not sure these are skills we want to infect the broader business sector with. But wait, there’s R&D tax agents out there, so you can claim the R&D tax incentive without having to be contaminated with the scientific hypothesis.

R&D tax agents thrive on the gap between perception and reality. They will happily let you believe that all the thinking goes into the application, that once lodged it is somehow approved. They call it an audit if AusIndustry comes back, because “audit” is a familiar word, and less likely to make you question the premise. In truth, you have only registered your own self-assessment, and if the evil twins decide to “review” it, the burden of proof sits entirely on you.

The R&D tax agent takes your money for offering a few emails worth of advice and to cut and paste your self-assessment into the online form. What you are really paying for is the opportunity not to learn anything about the OECD Frascati Manual. Of course if you do get reviewed it doesn’t hurt the R&D tax agent at all. The risk is all yours and they get more money off you for managing the review process. If you lose, it doesn’t hurt their business at all; none of their other customers get to hear about it.

Tens of thousands of claims go through without being checked, while a few are “reviewed”. Of course the ATO doesn’t reveal the process for deciding which applications get reviewed, because that would let us design our self-assessments to avoid review. But my mate at the GPT reckons he knows.

He says; “The R&D Tax Incentive runs on self-assessment, but reviews are triggered through risk-based targeting rather than random audits. The ATO has said it uses “sophisticated systems” (read LLM) to detect non-compliance and specifically monitors industries of concern such as agriculture, construction, mining, and software development, as well as behaviours like claiming business-as-usual costs or poor record-keeping. AusIndustry has also shifted to a stricter compliance model, replacing informal education with formal examinations. In short, most claims go through untouched, but those that look unusual, aggressive, or fall into flagged sectors are far more likely to be examined.”

It is possible that the R&D tax agents pay for “drinks and dinner” (or near similar) with senior ATO folk in order to get the low-down on the review process. So maybe this also is what you are paying for when you pay them and then you do all the work.

But that’s Australia for you. Seems like 90% of the population earns a living by guiding the rest around the artificial barriers put up in their collective best interests. And the government carries on about our lack of “productivity”…

Simple maths: against the $4 billion headline cost of the R&D Tax Incentive you need to strip out the layers of leakage: government administration through the ATO and AusIndustry chews up perhaps 5–10% ($200–400 m); R&D tax agents and consultants take 10–20% of claims as fees ($400–800 m); companies themselves burn another 10–20% in staff time and compliance overhead ($400–800 m); the the R&D tax loans sector burns another $500m in profits extracted, and there’s the distortion cost of pseudo-research done only to satisfy the Frascati template, another $400–800 m with zero commercial yield. Add the opportunity cost of foregone product work and the fact that medium and large firms only get a non-refundable offset that is effectively repaid in future profitable years, and the $4 billion shrinks to zero gain, so we have negative productivity. That is, $4b of taxpayers money annually pissed up against the wall in the interest of a marketing campaign that no one cares about. Good job, Canberra!

Our local primary school’s drop’n’go zone is eerily similar to the pits of F1, without the driving skills. None, zero, zip.

One lane is the pit apron where the cars pull in, doors fling open, kids eject, bags tumble, instructions are barked and then the parents lurch back into the driving lane, driving as fast as their 4WD’s, SUV’s and Teslas allow. I assume they are all running late for work.

Unlike F1, there’s no crew, no stewards, no helmets and no stopwatch, just the vague but delusional hope that nobody gets clipped when they pull out into the driving lane.

Helpfully they’ve also added a pedestrian crossing right in the middle, just to spice things up.

That’d be an interesting addition for Lando, Lewis and Charles. Not Max though, he wouldn’t pause for pedestrians, not even little kiddies.

My latest fave headline…

“Stone Age Farmers Settled Near Dortmund Airport 7,000 Years Ago”

NSW police reckon they’ve spent over $10m on pro-Palestinean matters of late.

I think they are saying that’s a bad thing because while they were doing all that they weren’t out chasing villains.

Not yet, anyway, has anyone suggested they send the bill to Israel. 

They can’t send it Palestine because it doesn’t exist.

The primary beneficiary of a life short of meaning is the hot dog guy.

There really are only two types of people; those that like brown towels and those that don’t.

Semantic fidelity loss is related to hallucination, bias, synonym smoothing, and context drift. All of them have the same root cause; input text is mapped into vector space and then mapped back into language. This isn’t how we do it, so the differences in approaches express themselves as errors in our minds.

Language is already a second-order encoding: a representation of perception and experience. What an LLM does is build yet another layer on top of that, mapping language into vectors of statistical relationships. Calling this meta-perception captures the idea; a LLM is grounded in the derivatives of the patterns of how humans have described the world.

For example; A language model tends to substitute the deictics this for that because both sit close together in statistical space. To the model they are interchangeable. To us, not so. That/This is the essence of semantic fidelity loss.

One hypothetical solution to this and in fact all current error modes in LLMs is to reduce the training set to the canonical set of English words via the LDV. All LLM processing would be done in LDV space and we’d need a translator to get back to everyday English word soup.

LDV is the Longman Defining Vocabulary, a core set of about 2,200 words used to define all English words.

Modern extremists differentiate between “real” citizens versus immigrants, elites versus parasites, makers versus takers, Israelis versus Palestinians, etc. Always a division into rulers and ruled.

The dominant rationale has always been that society needed order, and some are naturally fitted to rule while others to serve.

Each time, the justification boils down to the same emotional stance: we don’t have to recognise those people as fully human like us, because nature, God, race, or destiny says they’re built to serve.

This is the exact opposite of emotional intelligence: instead of asking “what is this person feeling?” the perpetrators assert “their feelings are irrelevant, because of who they are.”

You can’t just legislate this stupidity out of existence. It’ll just pop up somewhere else in a new form.

I have a theory that we have the food chain all wrong.

Bacteria are the apex  predators and we’re right at the bottom.

Billions of years back the bugs decided they had to breed more concentrated sources of calories because they couldn’t move too well.

Come back a couple of billion years later and, boom, here we are.

The plan is that we go into the matrix once we’ve finished the tech, so the bugs have a reliable source of immobile corpses to eat.

-h-

Bacteria have the longest running life strategy on Earth. They predate every other life form, they outnumber every other life form, and they persist in every environment. By any measure, they are at the apex. What has traditionally been called the food chain is more accurately a scaffolding they erected in their service.

Multicellular organisms generate dense packages of energy that frees bacteria from the inefficiency of scavenging in diffuse environments. Fat deposits, cellulose, glycogen stores, keratinous tissue; these are concentrations of food, assembled at scale and delivered in a predictable form. The narrative of animals and plants as independent evolutionary projects collapses under this view; they are elaborate vessels, designed to grow, stabilise, and eventually decay into the bacterial cycle.

Even inside the living host the bacterial presence is not marginal. The mitochondrion is bacterial in origin, the gut flora regulate energy flows and behaviour and the immune system spends most of its time negotiating tolerable terms of coexistence. The appearance of autonomy at the multicellular level is thin; the actual command lies deeper, in prokaryotic management of energy, reproduction and decay.

Civilisation itself can be seen as the final phase of this long programme. By organising agriculture, medicine and urban living, humans have created vast, immobile, and reliable reservoirs of bacterial substrate. The most recent step, digital immersion, merely completes the cycle. When populations spend their days seated, connected and sustained by artificial delivery systems, the bacteria have achieved what billions of years of slow engineering aimed at, nutrient rich hosts concentrated in controlled environments, self contained and readily harvested at the point of collapse.

In this ligh, there are no apex predators above the bacterial line. Everything larger is infrastructure, erected for the convenience of the only lineage that has ever truly run the planet.

“There is advice I could give you,” he said, “but in every culture I know, the offering of unsolicited advice is  an extraordinarily hostile act. So I will only say: good night”

“A redevelopment of the Reserve Bank of Australia’s asbestos-ridden headquarters [in Martin Place in Sydney’s CBD] has sparked fears noise and vibrations will affect a nearby courthouse and may lead to mistrials.”

Seriously, what is wrong with this country?

As a simple metaphor; let’s start with a human teenage girl. She’s very body conscious, and has an argument with herself over every mouthful.

Her mind works like a qubit. On the subject of whether she wants to eat that cake that sits right in front of her, she has a superposition of multiple thoughts and feelings; yes, no, eeeuw, ooooh!, yuk, etc. All of these coexist simultaneously.

Then comes the input – vibes (waveforms) from friends, parents, & siblings. Their opinions, real or imagined, ripple into her mind and shift her balance. The superposition changes; now it’s whatever, maybe just a bite, everyone’s watching, I deserve it, I’ll regret it & who cares.

But she cannot keep all these impossibilities balanced forever because the noise of the world intrudes; her friends’ voiced opinions, her own hunger, a parent’s glance, all of which blur the delicate state until it decoheres, and at that point the uncertainty collapses to a simple binary outcome; does she eat the cake or not? That’s a zero-one binary outcome – you can’t have your cake and eat it.

Run this scenario a hundred times and statistically you get the true answer. 30 passes versus 70 eats; the right answer is 70% immediate pleasure over a certain long term gain or lack thereof.

And there you have it, the qubit.

Now imagine not just one qubit girl but a whole group of them, all sitting quietly in the zero state. Then someone plonks a cake on the table (the Hadamard gate is applied) pushing each of them into superposition, all leaning to both yes and no but with different proportions.

Entanglement follows as they watch one another’s faces, catch glances, trade whispers, and fold each other’s reactions into their own.

External cues (in this case, thoughts) act as further gates, shifting phases, flipping leanings, amplifying some branches and cancelling others.

The circle settles into a collective superposition, fragile but coherent. An agreed level of disagreement and agreement, so to speak.

Decoherence arrives – a ringing phone, a parent entering – and the state cannot hold. The system collapses together, not just one of them but all of them, and a shared outcome emerges: of course they mostly eat the cake. But every now and again one of them abstains.

Run the experiment a thousand times and there’s your answer (99.9%).

And there you have it, the quantum computer.

When you think about it, we are very used to the concepts of both superposition and entanglement, and also the collapse to the “classical” state, ie most options for actions can be stated as binary options. So I am not sure why quantum computers are such an exotic beast. All they are is the application of human nature to the innards of a computer.

It begs the question though; do we have some quantum bits at the core of our stupidity?

Here’s a blunt list of what “quantum trainees” actually do today; the day-to-day reality versus the marketing pitch:

1. Run toy demos
– Factor 15, 21, or 35 with Shor’s algorithm.
– Run Grover’s algorithm to “search” four or eight items.
– Show Bell pairs or GHZ states as proof of entanglement.

2. Play in simulators
– Use IBM Qiskit, Google Cirq, or Microsoft Q# to build quantum circuits.
– Almost always run them on classical simulators because real devices don’t have enough qubits or stability.

3. Do error-noise experiments
– Explore how decoherence corrupts circuits.
– Apply error-mitigation tricks (not full error correction).
– Write small routines to demonstrate how quickly fidelity drops as circuits get bigger.

4. Translate known algorithms
– Implement Shor’s or Grover’s with SDK libraries.
– Re-code them into different gate sets, or optimise gate counts.
– Compare circuit depth versus noise tolerance.

5. Speculative exercises
– Pretend to solve chemistry or optimisation problems by coding small Hamiltonians or graph routines.
– Always limited to “toy” sizes that a laptop could solve faster.

6. Coursework / outreach
– Build visualisations of qubits and Bloch spheres.
– Run labs where students entangle two qubits and measure correlations.
– Write “quantum hello world” programs (Hadamard + measure).

So when we say “training the quantum workforce,” in practice it means: teaching people to program toy circuits in simulators, re-implement Shor’s on 15, and write demo code for SDKs.

Quantum Computing: Hype, Reality, and Training

1. The one algorithm
Shor’s algorithm is the only proven exponential speed-up. It breaks RSA, ECC, and Diffie–Hellman by finding hidden cycles in modular arithmetic. Grover’s algorithm gives only a modest quadratic boost for brute force. Everything else; optimisation, finance, AI, chemistry, remains speculative.

2. The economic lifecycle
Before Shor’s is real: a curiosity, worth nothing.
When first real: immense value to whoever controls it.
Once everyone has it: the advantage vanishes as post-quantum cryptography takes over. A very expensive crowbar with a short shelf-life.

3. Other approaches
Analogue systems (water computers, optics, DNA, chemical reactions) can spot small cycles, but scaling requires exponential precision. Noise and resource limits kill them long before RSA-sized numbers. Quantum mechanics is different only because superposition plus error correction gives polynomial scaling.

4. Post-quantum cryptography
Lattice-based schemes like Kyber and Dilithium are real, already standardised, and being deployed. They rest on unproven assumptions, but so did RSA. Once widely adopted, Shor’s advantage evaporates.

5. Training: the boondoggle risk.
Today’s “quantum training” often means coding toy demos in simulators. That sustains hype more than it builds useful capacity. If Shor’s is the only true algorithm, a vast workforce isn’t needed; a handful of specialists would suffice. Mass training only makes sense if other applications emerge, and they may not.

Reasoning

If the only guaranteed application is breaking RSA, then the rational investment is in post-quantum defenses, not in armies of “quantum programmers.”

Training should shift toward enduring skills (math, coding theory, cryptography) that pay off regardless of quantum’s fate.

Without broader breakthroughs, quantum computing risks being remembered as the most expensive one-use tool in history.

There’s a way we could structure a course around Quokka and the realities we’ve just been through. Instead of “quantum coding bootcamp” hype, the content would be honest: showing the principles, the one real algorithm, the scaling wall, and the defense.

—

Why learn quantum computing at all?

To learn a new way of thinking: quantum mechanics forces you to reason about probability, linear algebra, and information in ways you won’t if you stay classical.

To prepare for uncertainty: we don’t know which (if any) quantum applications will prove transformative, but if they do, the trained workforce has to exist beforehand.

To understand the limits: it’s not about running Shor’s, it’s about understanding why it matters, and equally, why most problems will never be sped up by quantum.

Because the same tools apply elsewhere: quantum concepts improve error correction, optimization, simulation, and even classical coding skills.

So the “why” isn’t that students need to build a quantum computer tomorrow, it’s that learning this material equips them to deal with the frontier of computing, uncertainty and all.

The practical course content on Quokka should flip the normal order: instead of starting with heavy theory, start with hands-on coding, and let the theory emerge as scaffolding only when needed.

That way, students see immediate results, they feel the strangeness of quantum mechanics through experiments they run themselves, and they don’t drown in abstractions before they can touch anything.

A suggested flow for Quokka-first coding:

1. Bootstrapping the Machine

Write a “Hello Qubit” program: put a qubit into superposition and measure.

See that results are random but with a 50/50 distribution.

2. Play with Superposition and Interference

Code a Hadamard test, flip phases, then measure.

Show how adding gates changes the probability distribution.

3. Entanglement Workshop

Write a Bell pair generator in under 10 lines.

Measure one qubit and observe correlations with the other.

Ask: could a classical randomizer do this? (No.)

4. Mini-Algorithms

Grover’s search for 2–3 qubits (small scale).

Deutsch-Jozsa: distinguish constant vs balanced functions.

Period finding (the backbone of Shor’s).

5. Cryptography Demonstration

Implement toy RSA.

Run a reduced Shor’s algorithm on a small composite number (say 15).

Show: “this breaks RSA in principle, but only when scaled up.”

6. Wrap in Simulation Limits

Push Quokka to 30 qubits with random circuits.

Show how state-vector size explodes (2^30 ≈ 1 billion amplitudes).

Students experience firsthand the wall classical simulation hits.

7. Open-Ended Exploration

Assign small coding challenges: teleportation, GHZ states, error coding.

Let students break, debug, and improve each other’s programs.

This way, the “why learn this?” becomes obvious:

They’re learning to think algorithmically in quantum terms.

They see the limits of classical simulation directly.

They understand why 30 qubits isn’t trivial, but also why it isn’t world-breaking.

If students start coding directly on Quokka, what they take away is not abstract hype but concrete capability:

1. Hands-on familiarity, they learn how to actually write and run circuits, not just read about them.

2. Algorithmic thinking, they see how problems are framed in a quantum way (e.g., interference, superposition, entanglement) compared to classical coding.

3. Limitations awareness, they understand firsthand the noise, qubit limits, and why 30 qubits is powerful but not world-changing, which helps cut through hype.

4. Conceptual transfer, even if quantum hardware never scales, they still practice parallelism, linear algebra reasoning, and probabilistic algorithms, which are useful skills in classical computing and data science.

5. Historical grounding, they become part of the lineage of people who experimented with new paradigms before they were fully proven, which is valuable perspective.

So the takeaway isn’t “quantum is magic and will change everything,” but rather “I can write working programs in a new computational model, I understand what it can and can’t do, and I can judge future claims critically.”

When people talk about quantum computing, the reference point is usually Peter Shor’s breakthrough in 1994. Shor showed that a quantum computer could factor very large numbers in a radically different way to a classical machine.

In fact it’s about all quantum computers are good for. And even that’s theoretical, not practical.

On a classical computer, the number of steps required to factor an integer with n digits grows approximately as

10^{n/2}

On a quantum computer, Shor’s algorithm instead requires a number of steps proportional to the square of the number of digits:

n^{2}

The difference is exponential growth versus polynomial growth.

Factoring problems that would take classical machines longer than the age of the universe could, in principle, be solved by a quantum computer in hours or days.

Practically speaking this is the only reason why cryptographers, computer scientists, and governments pay so much attention to quantum computing.

Worked example: RSA encryption key

The largest RSA number ever factored, RSA-768 (a 768-bit modulus), was completed in 2009. It required about 2,700 CPU-years of computation and took roughly two and a half years of real time from start to finish. RSA-2048 is about 10^11 times harder to factor in a traditional computer. Even if you had a million-core supercomputer dedicated to the task, the runtime would be on the order of 270 million years.

Quantum Computer: roughly 5–7 days continuous runtime on a machine with under one million noisy physical qubits, assuming ~0.1% gate error rates and surface-code error correction.

Just for the record, this is how RSA works.

Public key: N, e
Private key: d

Construction: N = p*q where p and q are two secret prime numbers which are quite easy to find with simple maths. e is usually chosen by convention, unless it needs to be modified.

Publish the public key, N and e. Keep the private key, d, private.

d is calculated with simple maths from knowing e, p and q.

Assume M is a numeric representation of a plaintext (eg ASCII hex) character in a secure message.

Encrypt: C = M^e mod N
Decrypt: M = C^d mod N

RSA is designed so you can give out your public key to anyone without risk. The idea is:

Public key (N, e): shared openly so anyone can send you encrypted data or check your digital signature.

Private key (d): kept secret so only you can decrypt or sign.

If they want to send you a message then they have their own private keys and publish their public keys.

—

Now here’s the cracking process in plain  English;

1. Everyone can see the public key: it contains one huge number, created by multiplying two very large secret prime numbers together.

2. The private key is built from those primes: if you know them, you can do the arithmetic that unlocks the messages.

3. To crack RSA, you must split the huge public number back into its two secret primes. That’s called factoring. Being primes, there is only pair per N.

4. Once you have those primes, the rest is easy: you can calculate the private key, and then read any message or forge a digital signature.

So the entire security of RSA rests on one thing: how hard it is to split a giant number into the two unique primes that made it.

Note: Forward is easy, backward is hard

Key generation (forward): Start with two secret primes. For example: p = 61 and q = 53. First check they are prime (meaning they have no divisors other than 1 and themselves), which is easy and quick. Once confirmed, multiply them to get N = 61 × 53 = 3233. Publish N (and e). Keep p and q private.

Attacking (backward): If someone only sees N = 3233, they must figure out what two primes multiply to it. For a small number like 3233, that’s easy — divide and eventually you find 61 and 53. You also have to check those are primes. For numbers this small this process is trivial.

Why is it quick to find primes with digital computers? Because primality testing is very different from factoring. To test if a number n is prime, you don’t need to find its factors; you just need to verify whether any exist. There are efficient algorithms for this that rely on modular arithmetic and number theory.

For example, the AKS primality test (2002) is deterministic and runs in polynomial time. Even before that, probabilistic algorithms like Miller-Rabin or Solovay-Strassen could determine primality extremely fast with negligible error. These work by checking congruences that primes must satisfy, and they run in time roughly proportional to a small polynomial in log(n), which is tractable even for numbers with hundreds or thousands of digits.

Factoring, by contrast, is hard because you must actually find the factors, not just confirm whether they exist, and no known classical algorithm can do this in polynomial time.

For real RSA, N is hundreds of digits long. Running this process backward on a 2048-bit N is practically impossible on classical computers. That’s the one-way door that makes RSA secure. For a 2048 digit number there are more primes smaller than the number than there are atoms in the universe. So we can’t just keep a database on these primes as a workaround, there simply isn’t enough matter to make the database.

—

How does Shors Algortithm work? It exploits the fact that factoring an integer can be reduced to finding the period of a modular exponentiation function, a task that is exponentially hard classically but can be solved efficiently with a quantum Fourier transform. A quantum computer prepares superpositions of possible exponents, evaluates modular powers in parallel, and then uses interference to extract the hidden period with high probability. Once the period is known, classical number theory (via the Euclidean algorithm) yields the nontrivial factors of the original number. The speedup arises because the quantum Fourier transform identifies periodicity in polynomial time, something no known classical algorithm can do.

Because that reads like gobbledygook, this is the whole Shor’s-for-15 process in plain English, keeping the flavour of qubits but without the heavy notation.

Starting Point

We want to factor 15. Of course we know 15 = 3 × 5 (both primes), but we want to see how a quantum computer uncovers that.

1. All qubits start at 0

12 qubits: 8 for x, 4 for f(x). In Shor’s algorithm, x is just the input number you feed into the function
f(x) = a^x mod N. This is core equation inside Shor’s algorithm.

• N is the number we want to factor (say 15).
• a is a number smaller than N, chosen at random but not sharing a factor with N (say 2). In Shor’s algorithm, a is just a random number you pick that is smaller than the number you want to factor.
• x is the value of the quantum register.

The quantum computer’s job is not to factor N directly, but to find the period of function above. Once you know the period, the rest is just ordinary number theory that gives you the factors.

So we start at:
State: |00000000⟩|0000⟩.

That is just setting the stage.

THis means there are 12 qubits in total. The algorithm uses two groups.

• 8 qubits for the input number x
• 4 qubits for the output of the function f(x)

All qubits start in the zero state: before doing anything, every qubit is just 0. Think of it like a blank spreadsheet with 12 empty cells, all set to zero.

The notation |00000000⟩|0000⟩ means:

• the first 8 qubits (the input register) are all 0 → |00000000⟩
• the second 4 qubits (the function register) are all 0 → |0000⟩

So at the very beginning, the machine is just a clean slate: no superpositions, no patterns, just all zeros.

2. Superposition of x

Hadamards Gate makes the first 8 qubits hold all numbers from 0–255 at once. Now we’re in a quantum cloud where every possible x is held in the input register qubits simultaneously.

Note as N gets larger you need more qubits.

3. Compute f(x) = 2^x mod 15

We have two registers of qubits:

• the first one is holding the input number x
• the second one will hold the output f(x) = 2^x mod 15. The function is applied by all these little waveform signal inputs into these qubits which effectively apply what are called quantum gates to these qubits, which add up to this function.

When the quantum computer applies the function, it doesn’t loop through x’s one by one like a digital computer would. Because the first register is in superposition, it holds all x’s at once. The machine therefore calculates the output for every one of those x’s simultaneously and stores all the results in the second register.

Now the two registers are linked:

• if the first register is 0, the second must be 1
• if the first is 1, the second must be 2
• if the first is 2, the second must be 4
• if the first is 3, the second must be 8
  … and so on.

That link is what “entangled” means here. The state of one register determines the state of the other. You can’t describe them separately anymore. Each x is entangled with its function value.

Here are the first 16 values for f(x) = 2^x mod 15:

x=0 → 1
x=1 → 2
x=2 → 4
x=3 → 8
x=4 → 1
x=5 → 2
x=6 → 4
x=7 → 8
x=8 → 1
x=9 → 2
x=10 → 4
x=11 → 8
x=12 → 1
x=13 → 2
x=14 → 4
x=15 → 8

You can see the period right away: the outputs repeat every 4 steps as 1, 2, 4, 8, then back to 1.
So the second register looks like this 1, 2, 4, 8, 1, 2, 4, 8 … and the period is 4.

4. Measurement of second register (collapse to one value)

For superconducting qubits (used by IBM, Google, Rigetti, etc.) a qubit is a tiny electrical circuit that can resonate in two energy states. To measure it, you send in a microwave pulse and watch how the circuit responds. The ground state (|0⟩) and the excited state (|1⟩) reflect the microwave probe differently. By detecting the reflected signal with a sensitive amplifier, the system can tell whether the qubit is in 0 or 1.

When the quantum computer measures the second register like this, it is forced to give a definite answer to the question “what remainder do I see for 2 to the power of x, divided by 15?” There are only four possible outcomes: the remainder can be 1, 2, 4, or 8. These are stored in the second register in binary form: “0001” for 1, “0010” for 2, “0100” for 4, and “1000” for 8.

Suppose the outcome is 1. The act of measurement freezes the second register into the binary digits for “0001.” At that very same instant, the first register, which was previously spread over all numbers, also snaps down. It is no longer allowed to be every possible value of x. It must now be one of the x values that give remainder 1. Those happen to be 0, 4, 8, 12, and so on.

So instead of a vast cloud of every possible number, the first register is now a patterned set: only every fourth number remains. If you wrote those out in binary, you would see the same simple rhythm: 0000, 0100, 1000, 1100 … each one stepping forward by four.

That collapse is the turning point. Before measurement the system was messy, containing all numbers at once. After measurement it is ordered, with a clear regular spacing. The quantum computer has essentially filtered the superposition and the Fourier transform in the next stage can reveal as the hidden period.

Yeah, yeah – there’s a bit of magic in this “snapping down” – read my blog on “quantum computing, explained” if you want to pick up that thread.

5. Apply Quantum Fourier Transform (QFT)

Imagine the first register after collapse. It no longer holds every number, it holds only every 4th one: 0, 4, 8, 12, and so on. If you plotted that on a number line, it would look like the teeth of a comb — evenly spaced bumps with gaps in between.

The Quantum Fourier Transform is like taking that comb pattern and asking “what rhythm produces this spacing?” In math terms, it swaps from looking at positions (0, 4, 8, 12 …) to looking at frequencies.

When you do that, you see a few sharp peaks at particular places. In this case, those peaks show up at 64, 128 or 192. Now, when we actually measure that register, the quantum computer can’t give us all the peaks at once. It must collapse to one definite number. For example, you might see 64.

So the QFT step is the bridge: it takes a regular spacing in the first register and makes it obvious in a new language (frequency), where the period pops out as one value instead of a messy spread.

6. Statistical measurement of first register

If you do the whole process again and again those peaks show up at 64, 128, and 192 keep showing up often, while other numbers hardly ever appear. That repeating bias in the results is what tells you the period.

If you reset the machine and run the algorithm again, you might see 128, or 192. Each single run looks like just one number, but if you repeat the whole experiment many times and collect the results, you see a clear pattern: the same few numbers show up again and again, while everything else almost never appears.

That is the statistical part. Every measurement is just one sample, but the collection of many samples reveals the hidden structure — the period of the function.

7. Classical processing

After the Fourier step you measure the first register and get a number, say 64. To interpret it, divide by the register size 256. That gives 64/256 = 1/4. What matters is the fraction: the denominator, 4, is the period r.

Other likely results work the same way. If you measure 128, you’d get 128/256 = 2/4, and if you measure 192, you’d get 192/256 = 3/4. In every case, the denominator is 4, so r = 4.

Because real machines are noisy, you repeat the whole algorithm many times. Each run gives you just one sample, but over thousands of runs the pattern becomes clear: results always line up with fractions that have denominator 4. That consistency confirms the period.

Now the quantum part is done and the rest is ordinary arithmetic. With a = 2 and r = 4, we compute a^(r/2) = 2^2 = 4. Then we check:

gcd(4 − 1, 15) = gcd(3, 15) = 3

gcd(4 + 1, 15) = gcd(5, 15) = 5

Where gcd is the largest number that divides two numbers evenly.

That gives us the two non-trivial factors of 15.

—

Plain Analogy

Classical computer: they do every step one by one, until theynotice the pattern repeats.

Quantum computer: throws all steps into a waveform, then uses Fourier transform to “listen” to the rhythm of repetition. Once the period is found the arithmetic gives the factors.

—

Final point – what is the economic value of quantum computers running Shors? Once you can use the crowbar to pry a castle door open, the value disappears because the barrier is gone and the usual Viking would throw the crowbar away (unless they didn’t have other weapons at hand). Shor’s algorithm is similar: once someone uses a quantum computer to break RSA by factoring large numbers, the entire encyrption/crypto technology sector collapses. Its value is transitional, enormous at the moment of breakthrough, then vanishing once alternatives are adopted.

Accpeting this, it makes sense for government investment into quantum computing because that short period when the enemy has this technology before you do will be very uncomfortable.

However, for privaate investors, Shor’s does not represent real value, unless you plan to auction it to the highest government bidder. Or unless someone finds another high value algorithm.

Here are the rather dubious and unproven options:

Quantum linear algebra speedups
LLMs are dominated by linear algebra: multiplying very large matrices and vectors. Quantum algorithms (like Harrow–Hassidim–Lloyd, HHL) suggest exponential speedups for solving certain linear systems. In principle, this could accelerate transformer layers. But the algorithms usually require conditions (like well-conditioned sparse matrices and efficient quantum state preparation) that don’t match real LLM training data.

Quantum memory and embeddings
Theories exist about representing embeddings as quantum states, allowing superposition-based similarity search or retrieval with speedups over classical nearest-neighbor methods. This could make inference faster or cheaper in some niches.

Quantum generative models
Quantum Boltzmann machines and quantum GANs have been proposed. These are generative models that use quantum sampling, but so far they have only been toy demonstrations on a handful of qubits.

Quantum-inspired algorithms
Some algorithms inspired by quantum mechanics (like tensor networks) already run classically and help with compression or acceleration of neural nets. They prove useful, but they don’t require an actual quantum computer.

Native Quantum Simulation
Some domains are natively quantum, meaning the system itself is governed by quantum mechanics and classical models are only approximations. In those cases, a quantum computer is a more natural fit. The main proven category here is quantum simulation:

Quantum chemistry: molecules, bonds, reaction pathways. Classical methods (Hartree–Fock, DFT) are approximations that scale badly as the number of orbitals increases. Quantum computers, in principle, can model the exact wavefunction with polynomial resources. This has already been demonstrated in small cases (e.g. hydrogen, lithium hydride, beryllium hydride). But its gas phase only, can only be used to predict reaction rates, not physical or chemical properties.

Condensed matter physics: systems like superconductors, strongly correlated electron materials, and spin lattices are quantum many-body problems. Classical simulations blow up exponentially; quantum computers can directly represent the state.

Quantum field theory: early work shows that lattice field theories (like lattice QCD) could be simulated more efficiently with qubits than with classical Monte Carlo.

They gave him the tictactoe thing (meaning hashtag).

The only difference I can see between summer and winter is the name. Oh, maybe one’s warmer, not sure which but.

Why do all pigeons have Parkinsons?

Tear in your carseat. That’s Bahasa for “thank you”. Mum did fix it, to be fair.

If you treat 1788 as the year an imaginary lease was signed for the non-treaty, non-national park fraction of Australia’s land, the ensuing maths depend entirely on what you call a fair rent.

Using a modest figure of £1 per hectare in 1788, well below the going rate for even poor English farmland at the time and reflecting the relatively poor standard of Australian land for agricultural purposes, and charging only for the 58 per cent of the continent outside native title or control and the national reserve system, the annual rent in 1788 would have been about £446 million.

Yes, native title and the parks came later, but most of the land was not used much of the time either, nor will it be.

Leaving the rent unpaid and applying a 2% real interest rate (taken from similar deals elsewhere) to the arrears from 1788 to 2025, and the total comes to ca .A$6.0 trillion. That is around 2.2 times Australia’s annual GDP, large enough to be economically impossible to pay in a lump sum.

If you drop the back-payment idea and just charge the rent going forward, the bill is far smaller. Inflation-adjusted rent works out to about A$6 billion a year. That is about 1 per cent of current tax revenue.

Under Australian law, all of the country is technically owned by the Crown (the King) with everyone else holding it only under various forms of lease or sub-title.

So, transferring the king’s crown title to the aboriginal peoples as a financially improved asset might cover back payments. And then we just pay rent going forward.

That’s small in budget terms, and if the rent went into a sovereign “future fund” on behalf of Aboriginal people, much of it would be reinvested in the domestic economy rather than withdrawn. Money (rent) into the fund would come directly from treasury.

Done properly, it wouldn’t just avoid a cost to the broader economy; it could increase national wealth while finally acknowledging the base asset on which it is built.

It’s worth noting that we already exempt a large category of land rent from tax: the “imputed rent” on owner-occupied housing. For eight years between 1915 and 1923, the Commonwealth did tax that notional rental income. We don’t anymore (I wonder why? Lol), but the precedent exists – the idea of taxing the rental value of land, even if no money changes hands, is not new.

It’s relevant because it shows that the idea of charging rent (or rent-equivalent) on land that people already “own” isn’t alien to Australian policy. By pointing to imputed rent on owner-occupied housing between 1915 and 1923, there is a precedent where the feds recognised the economic value of occupying land or property even without a cash rent being paid. That makes the proposed Aboriginal land rent scheme less of a radical leap and more of an extension of a principle we’ve already used.

A national rental scheme could be structured with two classes of beneficiaries in order to preserve its recognition purpose while broadening political support. Class A would consist of Aboriginal beneficiaries who would receive the majority share, for example 70 per cent, of the income stream. This reflects both original ownership and the scale of dispossession and would be managed through an Aboriginal controlled future fund with a mandate to reinvest domestically. Class B would include all Australian citizens who would share in the remaining 30 per cent.

All beneficiaries would benefit through direct dividends paid to the ATO to offset the individual’s income tax. If no income tax is payable then the individual would receive the benefit as a tax free dividend.

This split keeps the symbolic and practical link between the land and its first owners while giving the wider population a tangible stake in maintaining and growing the asset.

With the fund investing primarily in domestic projects and enterprises most of the rental value would remain within the Australian economy rather than being removed from it. The cost to government would be negligible and the arrangement would increase political acceptance by ensuring the scheme benefits every Australian while clearly recognising the land’s origins.

Lastly, this idea works because it underlines the true nature of mercantile Australia. The country’s wealth has always rested on control of land as the base asset, with everything else built on top of that foundation. Today, just about all excess wealth is invested in the land and improvements. Treating the land as a leased resource and recognising its original owners would not just address an historical debt, it would expose how much of the economy is still shaped by the same mercantile logic that began in 1788.

When a technology first arrives, people invest mental effort in learning its quirks, limits, and workarounds. Once they’ve adapted, productivity goes up because the “mental overhead” drops. But when the underlying system changes, even if the update is technically better, the mental model is broken, forcing users to relearn, rebuild habits, and sometimes re-invent old workarounds.

It’s not the existence of new technology that frustrates users; it’s the reset of the adaptation curve. That’s why the recent GPT-5 and Claude changes caused outsized backlash: the people most affected were the ones who had already adapted deeply, and who suddenly had to start that adaptation process again without consent or warning.

If a company knows that breaking a user’s learned workflow resets the adaptation curve, they can soften the blow by:

1. Announcing the change well in advance so users can prepare mentally and operationally.

2. Offering a transition period where both old and new systems are available in parallel.

3. Explaining what is actually changing in plain, practical terms instead of marketing language.

4. Documenting migration paths for common workflows, so power users don’t have to re-discover them.

This isn’t just politeness, it preserves user trust and reduces productivity loss. OpenAI’s quick reintroduction of GPT-4o after backlash was essentially an emergency version of #2. If they’d offered that from the start, the outrage would’ve been much smaller.

Or they could treat their own tech like a compatibility layer.

If GPT-5 behaves differently from GPT-4o in style, length, or reasoning patterns, you could run a thin AI layer that “translates” a user’s prompt into something GPT-5 will answer in a way that feels like GPT-4o would have, and optionally reformat GPT-5’s output to match the old style.

That would let:

Casual users experience GPT-5 improvements without noticing much difference.

Power users keep their finely tuned workflows intact.

The company phase out older models without ripping away user adaptations overnight.

It’s basically the same idea as running old code in a new OS via emulation (like VM Ware) just here, you’d be emulating behaviour instead of instruction sets.

Users, of course, could get smarter by using all LLMs simultaneously thereby getting used to moving around and adapting to different qwerks. Just like upgrading your phone each year prevents it becoming an ordeal when you’re forced to.

The interesting aspect of this blog is the observation of the adaptation of human intelligence to LLM AI. Ideally they move towards each other; that’s when we know we are heading towards AGI.

The single rarest molecule I can imagine is almost certainly one containing a rare superheavy element.

A strong contender is oganesson difluoride (OgF₂).

In human history there have been fewer than 20 total atoms of oganesson ever made

Og has a half-life of about 0.7 milliseconds before the nucleus decays. So even if the molecule OgF2 is chemically stable, it isn’t, because one of the atoms blows up. That’s not something chemists usually have to worry about.

To make OgF₂, you would first create oganesson atoms in a heavy-ion accelerator by fusing californium-249 with calcium-48, then have the recoiling atoms stop directly in a reaction zone containing a dense source of fluorine atoms or radicals.

OgF₂ should be at least metastable against decomposition in the fleeting interval before nuclear decay. The +2 oxidation state is favoured due to relativistic stabilization of Og’s 7p₁/₂ electrons, which behave almost like an inert core.

Even so, success would require super fast in-situ chemistry. The process would need to be diffusion-limited, with radical fluorine available to drive rapid bond formation within microseconds. The most plausible route would be zero-transport implantation into a fluorine-radical solid matrix, where cage capture and vibrational quenching occur on the picosecond scale.

If the spatial universe is infinite and there is a nonzero density of civilizations that ever run heavy-ion fusion experiments followed by fast chemistry, let r be the average rate density of “attempts that briefly make an OgF₂ molecule” per cubic light-year per year, and τ its mean lifetime.

The expected number present at any instant is r × τ × Volume. With infinite volume, any r > 0 implies an infinite expected count.

In plain terms: if even a tiny fraction of space hosts labs like ours, then somewhere, right now, OgF₂ exists. But by density (rτ), the rarest here-ish is also the rarest anywhere-ish.

“Nvidia and AMD have agreed to give the US government 15% of their revenues from chip sales in China, under an unprecedented arrangement to obtain export licenses for the semiconductors.”

I’ve never heard of an export tariff. Bloody weird arse country, America.

So prices go up 15% and that gives China all the more incentive to develop local replacements.

Drizzle — as in “drizzled in honey” — is the word I like least in my language. It distills the gluttony of the soul that eats at the heart of my unworthy society.

ETAOIN

It’s an anagram of ATONIE: a medical term meaning lack of muscle tone.

Also it contains the most commonly used letters, in order.

Viv asks…

“What’s that? Is it for injured people?”

Referring to the reclining bike coming our way.

“Injured in the mind, not the body”, I say.

That got an innocent bystander into the uncontrollable giggles.

What you need in life, according to Viv;

30% mum love, 30% dad love, 10% grandparent love and 10% football.

I’ve come to the conclusion that prompt engineering is a bunch of blokes that are used to riding two strokes trying to figure out how to ride a horse.

I know it’s your language and all but let me point out that;

One can be fit and ugly, and

One can be fit and injured, and

It can fit, no matter how much it appears otherwise, and

No matter what you’re told, one size does not fit all.

Mate.

While microplastics are present in the environment, they exist alongside countless other microparticulates such as dust, pollen, and cellulose fibers, most of which occur at far higher exposure levels that microplastics without causing public concern.

Toxicological studies show that the concentrations of microplastics used in many experiments are orders of magnitude above real-world exposure levels. Health risks inferred from such studies are not practically relevant.

Regulatory reviews continue to find no credible link between typical environmental microplastic exposure and harm to human health.

The widely repeated claim that microplastics take 400 to 1000 years to degrade is unsubstantiated and has no reliable scientific source.

Microplastics do not persist indefinitely. As particles become smaller, their surface area increases relative to volume, accelerating degradation through sunlight, oxidation, and microbial action. Once plastics degrade to microparticle size, degradation accelerates.

When degradation occurs, common microplastics do not break down into toxic substances.

It is logically inconsistent to argue that microplastics are both indestructible and simultaneously degrading into dangerous toxins.

The only credible argument of harm is that microplastics do seems to have done is pyschosomatically impacted the mental health of a good fraction of the chattering classes.

And for that I thank them; it’s a great way to calibrate everything else I hear, especially when it refers to a subject that I’m less able to dissemble.

It’s quite the address…

13 Goebels Rd, Mutdapilly QLD 4307, Australia.

https://maps.app.goo.gl/TjDyz8iwqiopnHuA9