Too lazy to submit.
Scientific ideas, half-baked hypotheses, and things that probably deserve a proper paper — but here we are. Physics, mathematics, AI, music theory, and education — explored with varying degrees of rigour. Peer review welcome. Criticism will be posted.
68 articles2020–2026 corrections via GitHub issues
2026
14 articles- The Day Claude Decided to Delete My Whole Codebase On 5 October 2026, a unit test run by a coding subagent recursively deleted every checkout under my projects directory in well under two minutes. No model “decided” anything. An orchestrating model wrote a guard rule of the form “anything outside the repository is allowed”, a test handed that guard the repository’s parent directory, and nothing between the test and the filesystem said no. This is the incident, the recovery, and the four layers that now stand where there were none.
- Token Saving as Noise Reduction Token saving in agent systems is often framed as an efficiency concern. A better framing is signal-to-noise control: every unnecessary token competes for attention, increases inference cost, and makes the next decision less precise. Current work on context compression, KV-cache compression, and long-running interactions points in the same direction.
- GitHub Stars Are for Sale, and the Real Problem Is Not Vanity Buying GitHub stars sounds pathetic enough to be a joke. The joke misses the interesting part. Stars are a public trust signal on a code-hosting platform, and there is an ordinary, low-friction market for faking them. The important question is not whether inflated stars reliably produce downloads. It is why counterfeit popularity is so cheap to buy, and why that becomes a security problem once fake credibility attaches to malicious repositories.
- There Is an App for That — Until There Isn’t Digital health applications and psychotherapy follow different capacity, evidence, and reimbursement paths. Baumol’s cost disease explains one pressure on labour-intensive care; it does not explain a particular German fee decision or establish that an app substitutes for a therapist.
- The Model Has No Seahorse: Vocabulary Gaps and What They Reveal About LLMs There is no seahorse emoji in Unicode. A language model may nevertheless claim there is one or emit a plausible substitute. That failure is real, but the output alone does not tell us whether tokenisation, training data, prompting, decoding, or verification caused it.
- Oppenheimer Didn’t Have an Acceptable Use Policy Anthropic has drawn a public line on military use of its models. The physics community spent the better part of the twentieth century working out what it means to draw that line after you have already built the thing. As a physicist watching this unfold, I find the parallels clarifying and the differences more unsettling than the parallels.
- More Context Is Not Always Better The intuition that feeding a language model more information improves its outputs is wrong often enough to matter. Here is why, and what to do about it.
- If You Think This Is Written by AI, You Are Both Right and Wrong AI detectors can flag human writing, including historical and highly formulaic prose. Evidence exists for a false-positive risk affecting some non-native English writers; claims about autistic or physics-trained writers need direct validation. The detector score is not authorship evidence.
- Two Routes to the Expansion Rate: The Hubble Tension Near 5σ A local distance-ladder measurement and the value inferred from the cosmic microwave background under ΛCDM disagree by about 4.9σ for one widely quoted comparison. JWST has tested an important Cepheid-crowding systematic, while DESI finds a dataset-dependent preference for evolving dark energy that does not by itself resolve the Hubble tension. The remaining possibilities include unrecognised systematics and physics beyond the standard cosmological model.
- Car Wash, Part Three: The AI Said Walk In early 2026, variants of a prompt about a nearby car wash spread online after some assistants recommended walking instead of taking the car. This is neither the tokenisation failure from the strawberry post nor the grounding failure from the rainy-day post. It is a pragmatic inference failure: the response followed the travel wording but missed the likely purpose of the trip, assigning its advice to the wrong interpretation of the question. A third and more subtle failure mode, with Grice as the theoretical handle.
- Automate the Boring Stuff: Setlist to Playlist I love concerts. I love setlists. I hate building the playlist manually afterward. But do I really? A small automation project, a Deftones show in Dortmund, and the question of whether you should automate something you kind of enjoy.
- Your Encryption Keys Are in Virginia: On BitLocker, the FBI, and Why European Universities Need Sovereign Software Microsoft confirmed last week that it produced BitLocker recovery keys to the FBI after a valid legal order. On eligible devices, Windows can back a recovery key up to a Microsoft account during setup. For European universities that handle research data, student records, and HR information under GDPR, this is not an abstract concern. It is a structural problem. The answer is not a technical workaround. It is sovereign, publicly funded, openly licensed software — and a principle that the EU has articulated but not consistently practised: public money, public code.
- Should I Drive to the Car Wash? On Grounding and a Different Kind of LLM Failure A viral video this month showed an AI assistant confidently answering “should I go to the car wash today?” without knowing it was raining outside. The internet found it funny. The failure mode is real but distinct from the strawberry counting problem — this is not a representation issue, it is a missing-context issue. The words posed no special tokenisation difficulty. What the system lacked was access to the state of the world the question was about.
- Try to Relax — and Other Things That Prevent Themselves “Try to relax” can recruit the monitoring that keeps tension salient. Traversable wormholes face a very different constraint: sustaining their geometry requires stress-energy that violates classical energy conditions. The comparison is an analogy about constraints, not one mechanism shared by psychology and spacetime.
2025
15 articles- A Christmas Star (Minus the Star, Plus a Moon Nobody Asked For) A browser-based simulator for exoplanet transit photometry, binary eclipses, and exomoon scenarios — with Keplerian and N-body paths, limb-darkening models, and timing diagnostics. It is an educational simulator, not a validated research pipeline. I spent Christmas on this. You’re welcome, science.
- The Golden Bead Cube Weighs One Kilogram Bruner described enactive, iconic, and symbolic forms of representation; Montessori materials make a useful, but not historically direct, comparison. Physical materials can support particular learning tasks, while evidence does not justify treating a touchscreen as a universal developmental failure.
- Constraining the Coding Agent: The Ralph Loop and Why Determinism Matters In late 2025, agentic coding tools went from impressive demos to daily infrastructure. The problem nobody talked about enough: when an LLM agent has write access to a codebase and no formal constraints, reproducibility breaks down. The Ralph Loop is a story-driven harness intended to bound that variability through one model invocation per attempt, scoped writes, and atomic state. These constraints make runs easier to inspect; they do not make an unpinned model deterministic.
- A Gas at Temperature T: Xenakis and the Physics of Stochastic Music Iannis Xenakis used probability distributions, Markov processes, and game structures as compositional tools. Pithoprakta maps a Gaussian law derived from kinetic theory to glissando slopes; the orchestra is an analogy to a gas, not a thermodynamic system with measured temperature.
- From Oxide to Oversampling: The Physics of Recorded Sound ‘Analogue warmth’ and ‘digital coldness’ are not aesthetic preferences alone — they involve different physics. Magnetic tape saturates and compresses in level- and machine-dependent ways. Delta-sigma modulators push much quantisation noise out of band. Both effects are calculable.
- Can the AI Friend Make You Lonelier? AI companions promise to address loneliness. Daniel Wegner’s ironic process theory offers one speculative account of why companionship use may be risky for some people; current studies do not establish that companions cause loneliness.
- Star Polygons and Drum Machines The {7/2} heptagram is not only a symbol. It is a traversal algorithm over seven beat positions. Because 7 is prime, that traversal never gets trapped in a sub-orbit.
- The Cat’s Eye: Slit Pupils, Thin-Film Mirrors, and 135-Fold Dynamic Range Domestic-cat pupils change area by about 135-fold, and a reflective tapetum sends transmitted light back through the retina. The pupil/ecology link is a comparative correlation; simple multilayer equations illustrate tapetal optics but do not supply cat-specific reflectance or sensitivity values.
- Your Transcript Is Already an Interpretation: AI Transcription and Grounded Theory aTrain and noScribe can run locally and reduce data-transfer exposure in tools that can genuinely save hours of work in qualitative interview research. They also make methodological decisions on your behalf without telling you. If you do grounded theory, you need to know which decisions those are.
- There Is No Blue Pill: The Epistemology of the Red Pill/Blue Pill Choice The red-pill choice is under-evidenced, not formally impossible to make rationally: its evaluation depends on priors, likelihoods, utilities, and unknown consequences. Cypher’s preference for simulated experience can be separated from his betrayal.
- The Oldest Algorithm in the World Plays the Clave Euclid’s algorithm for computing greatest common divisors, applied to the problem of distributing k drum beats as evenly as possible among n time slots, produces patterns that can be compared with timelines documented in several musical traditions. The mathematical similarity does not establish a shared origin, cultural equivalence or historical transmission.
- The Papertrail: AI PDF Renaming and the Tokens That Make It Interesting Everyone has a Downloads folder full of “scan0023.pdf” and “document(3)-final-FINAL.pdf”. Renaming them by content sounds trivial — read the file, understand what it is, give it a name. The implementation reveals something useful about how LLMs actually handle text: what a token is, why context windows matter in practice, why you want structured output instead of prose, and why heuristics should go first. The project, now called Folionym, is at github.com/sebastianspicker/folionym.
- The Oracle Problem: What The Matrix Got Right About AI Alignment The Oracle is the most interesting character in The Matrix for anyone who thinks about AI alignment. On one reading, she manages Neo’s beliefs for an outcome she judges good. That makes her a useful fictional case about honesty, autonomy, and paternalism—not a textbook diagnosis of a real AI system.
- From Thought Experiment to Qubit: Schrödinger’s Cat at Ninety In 1935, Schrödinger introduced the cat as a reductio ad absurdum of quantum superposition. Ninety years later, “cat states” — superpositions of coherent states with opposite phases — are a practical tool in quantum computing. Bosonic cat qubits can bias bit-flip errors exponentially with photon number under appropriate stabilisation, and are a serious candidate architecture for fault-tolerant quantum computation. The cat is no longer a paradox. It is a qubit.
- Below Threshold: What Google’s Willow Chip Actually Proved In December 2024, Google published a Nature paper announcing that their Willow chip demonstrated quantum error correction below the threshold — the point at which larger codes become more reliable, not less. The headline about “10^25 years of classical computation” was technically true and mostly a distraction. The real result is more important and less flashy: for the first time, a quantum processor demonstrated that logical error rates decrease exponentially as the code grows. This is what scalable quantum computing looks like at its first credible step.
2024
18 articles- Artificial Intelligence in Music Pedagogy: Curriculum Implications from a Thementag On 2 December 2024 I gave three workshops at HfMT Köln’s Thementag on AI and music education. The handouts covered data protection, AI tools for students, and AI in teaching. This post is the argument behind them — focused on the curriculum question that none of the tools answer on their own: what should change, and what should not?
- Inner Echo: On Making Mental Illness Visible, and What That Even Means I am on the spectrum. This post asks what “making mental illness visible” can realistically mean and describes inner-echo, an unvalidated audiovisual metaphor project—not a simulation, diagnostic instrument, or therapy.
- After the Connection Is Stable, the Hard Part Begins A third post in the networked music performance series. Technical latency and institutional infrastructure are only part of the problem. This practice-informed account describes one curriculum’s use of progressive tasks, reflection and portfolios without treating observations as measured learning effects.
- Primes Are Energy Levels: The Montgomery-Odlyzko Conjecture In October 2024, the largest known prime was announced — 41 million digits, found by a GPU-assisted GIMPS search. But the deepest prime story is not about record-breaking numbers. It is about a 1972 teatime conversation at the Institute for Advanced Study, a pair correlation formula, and the suspicion — numerically confirmed to extraordinary precision — that the zeros of the Riemann zeta function are the energy levels of an undiscovered quantum system.
- The Hamiltonian of Intelligence: From Spin Glasses to Neural Networks On October 8, 2024, Hopfield and Hinton were awarded the Nobel Prize in Physics. The Hopfield energy has Ising form, and Boltzmann machines import statistical- mechanical distributions. Those are genuine mathematical connections, but they do not make transformers, AlphaFold, protein folding, and spin glasses one causal lineage or universality class.
- Three Rs in Strawberry: What the Viral Counting Test Actually Reveals The viral “three Rs in strawberry” test exposes a real mismatch between token-level input and character-level operations. It does not, by itself, identify one tokenizer boundary or explain any particular model response.
- Why Cats Purr at 25 Hz: Vocal Fold Pads and the Physics of Self-Sustained Oscillation A 2023 Current Biology experiment showed that excised cat larynges can produce purr-like low-frequency oscillations under airflow without neural input. Connective-tissue pads are a plausible contributing mechanism, not a settled complete account; frequency overlap does not establish bone healing in cats.
- The Invisible Entrance Fee: On Privilege, Education, and the Institutions That Reproduce Both Education is supposed to reward merit, yet German PISA outcomes remain strongly associated with socioeconomic background. Bourdieu’s theory of capital offers one account of how institutions can reproduce advantage; it is an interpretive framework, not the only causal explanation.
- Why 44,100? The Accidental Physics of the CD Sampling Rate The CD sampling rate grew out of 1970s video-based PCM recorders. PAL-compatible equipment used 44.1 kHz; colour NTSC equipment used the nearby 44.056 kHz. Sony and Philips then chose one rate for the disc: 44.1 kHz.
- How Cats Drink: Inertia, Gravity, and the Froude Number at the Tip of a Tongue Cats use the dorsal surface of the tongue tip to entrain a liquid column, then close the jaw before gravity and pinch-off remove it. Reis and colleagues modelled the timing with inertia, gravity, a Froude number, and the tongue’s travel geometry. The relevant domestic-cat parameter is a modified Froude number near 0.4—not the simple Froude number near one.
- The Boring Parts of Networked Music Performance A follow-up to the August 2023 latency post. The numbers were fine. The hard part turned out to be everything else: governance, maintenance, invisible labour, and why a technically functional Digital Music Lab can become unusable without sustained stewardship.
- There Is No Such Thing as Full Accessibility — Only Barrier Reduction The German word ‘Barrierefreiheit’ promises freedom from barriers. That promise is structurally impossible. What we can achieve is Barrierearmut — a reduction of barriers. The difference is not semantic; it has consequences for policy, design, and institutional honesty.
- Are Cats Liquid? The Deborah Number and the Rheology of Cats Marc-Antoine Fardin won the 2017 Ig Nobel Prize for a playful application of rheological language to cats. The Deborah number De = τ/T is a standard timescale ratio for materials; assigning a cat a relaxation time is an analogy, not a rigorous material classification.
- Hunting Exoplanets with Your Phone: A Classroom Experiment That Actually Works Finding planets around other stars sounds like it requires a space telescope. It does not — at least not the analogy version. This is the story of how a lamp, a ball, and a smartphone became a peer-reviewed physics classroom experiment, published in The Physics Teacher in 2024.
- You Cannot Have All Three: The Fairness Impossibility Theorem For an imperfect binary classifier with unequal group base rates, predictive parity and equal false positive and false negative rates cannot all hold simultaneously. This is not an engineering failure. It is a theorem. Choosing which criterion to prioritise is a normative decision, not one made by mathematics alone.
- What the Videography Manual Didn’t Cover: Filming Music Education The classroom videography manual we published in 2023 was about filming teaching. Music education has the same word in it — teaching — but it is a fundamentally different recording challenge. Sound is the subject matter. The lesson is often one person, in a practice room. And the feedback cycle the teacher needs to reach is mostly the one that happens when no camera is present. A reflection on what the manual missed, and a software prototype that tries to address part of it.
- When Musicians Lock In: Coupled Oscillators and the Physics of Ensemble Synchronisation The Kuramoto model explains how heterogeneous phase oscillators can develop collective synchrony. It is a useful analogy for ensemble timing, applause, and delayed feedback, but it does not by itself predict a universal networked- music latency limit or prove neural coupling between musicians.
- The Impossible Heptagon A regular heptagon is not constructible by classical compass and straightedge. That theorem can illuminate seven-fold imagery, but it does not document Danny Carey’s compositional or symbolic intent.
2023
11 articles- Twelve Is Not an Accident: The Group Theory of Musical Tuning Why has twelve-tone equal temperament become so useful? Continued fractions explain its unusually good fifth; cyclic-group structure explains some pitch-class symmetries. Neither fact makes it the only musical possibility.
- Non-Commutative Pre-Schoolers Socks before shoes is a useful analogy for order-dependent operations. It does not explain quantum uncertainty, and infant sequence memory is not evidence of group-theoretic understanding. The shared structure is narrower: changing operation order can change, or prevent, the result.
- LK-99: Six Weeks That Showed How Physics Works On July 22, 2023, a Korean preprint claimed that LK-99 — a copper-doped lead apatite — was a room-temperature, ambient-pressure superconductor. Within six weeks, converging null results and impurity studies had rejected the claim. The episode is more interesting for what it revealed about the sociology of science than for the compound itself: how a global community self-corrected at extraordinary speed, and how the media managed to fail at conveying uncertainty despite watching it happen in real time.
- Zero Angular Momentum: The Falling Cat and the Geometry of Shape Space A cat dropped upside-down rotates 180° and lands on its feet, despite having zero angular momentum throughout. This is not a trick and not a violation of physics. The explanation took physicists from 1894 to 1993 to fully work out, and the answer — a geometric phase arising from the holonomy of a fiber bundle — is the same mathematics that governs the Berry phase in quantum mechanics and the Aharonov-Bohm effect in electrodynamics. We adopted two strays this year. They fall beautifully.
- Can a Planet Have a Moon? Teaching Exomoon Detection with a Disco Ball Motor Every classroom treatment of exoplanet detection focuses on the transit method. What gets omitted is that moons of exoplanets could also host life — and that with a small motor and a slight modification to the standard transit experiment, you can show students what an exomoon signature looks like in a light curve. Published in MNU Journal in 2023.
- How Low Can You Go? Measuring Latency for Networked Music Performance Across Europe We measured end-to-end audio and video latency for LoLa and MVTP across six European research-network links. One-way audio latency ranged from 7.5 to 22.5 ms in these deployments. Route topology and endpoint processing made air distance alone a poor predictor; one firewall comparison found higher loss and buffer instability.
- The Milky Way Is a Gravitational Wave Detector LIGO uses 4-kilometre laser arms to detect gravitational waves at hundreds of hertz. Pulsar timing arrays use millisecond pulsars scattered across the Milky Way — arms measured in light-years — to detect gravitational waves at nanohertz frequencies, ten orders of magnitude lower. In June 2023, four collaboration papers reported evidence for a stochastic gravitational wave background. The Milky Way itself was the detector.
- Fremde Welten: Teaching Exoplanet Detection in the Secondary School Classroom A unit for lower secondary physics classes (grades 8–10) on detecting exoplanets with analogy experiments. Published in Unterricht Physik in 2023, it starts where students’ misconceptions are — with the (wrong) assumption that you can just look at exoplanets through a telescope — and works forward from there.
- How to Actually Film a Classroom: An Open-Access Manual on Classroom Videography Three years after writing about why classroom video works, Charlotte Kramer, Kai Kaspar, and I wrote a manual on how to actually do it. The gap between interpreting study-specific evidence for video-based learning and producing usable footage turns out to be substantial. The manual is open access. Here is what is in it and why some of it surprised me to write.
- The Charm of Impossibilities: Group Theory and Messiaen’s Modes of Limited Transposition Messiaen’s seven modes of limited transposition cannot be fully transposed through all twelve keys — not by convention, but because of group theory. The modes are pitch-class sets whose stabiliser subgroups in ℤ₁₂ are non-trivial. The orbit–stabiliser theorem gives the exact count of distinct transpositions for each mode. The subgroup lattice constrains the possible symmetry types; it does not select Messiaen’s seven scales.
- Nobody Is Normal: A Roast I Had to Retract I wanted to roast the psychiatric slogan “Up close, nobody is normal” for trivialising severe illness. The evidence does not support that verdict. Continuum beliefs are generally associated with less stigma, although a slogan alone is not a tested intervention and clinical thresholds still matter.
2022
5 articles- Spiral Out: Tool’s Lateralus, the Fibonacci Sequence, and the Mathematics of Musical Structure Alongside physics and astronomy, two other things have occupied an unreasonable share of my attention since adolescence: mathematics and music. Lateralus by Tool — released 2001, still in rotation — is the piece that most conspicuously occupies the intersection. The song is frequently analysed through Fibonacci numbers, including reported lyric counts and a 9/8–8/8–7/8 metre grouping. This post separates checked arithmetic from listening analysis and authorial interpretation.
- What Black Hole Images Actually Show (and Why a Wormhole Would Look Different) The EHT images of M87* and Sgr A* are remarkable not because they surprised us, but because they confirmed a century-old prediction at microarcsecond precision. The more interesting question: what would a wormhole look like? Completely different — and we have never seen that.
- The Lab Goes Home: astro-lab@home and the COVID Pivot in Astronomy Education In spring 2020, the astro-lab at the University of Cologne shut down like everything else. The question was whether you could replicate a hands-on student lab using smartphones and household materials — and send it home. This is the story of how we tried, what we published in CAPjournal, and what happened when schools reopened.
- Why Universities Need Their Own YouTube In June 2022 I presented on educast.nrw at the Tag der Lehre at HfMT Köln. This is the longer argument behind that talk: why universities should not outsource their video infrastructure to commercial platforms, and what a better alternative looks like in practice.
- Teaching Stellar Evolution Without a Star: DIY Experiments and a Board Game Stellar evolution is now in the NRW physics curriculum, but there are almost no direct experiments you can do with it. Two responses: some DIY smartphone experiments for stellar formation, and a board game called “Staub und Sterne” (Dust and Stars) that lets students play through the stellar lifecycle. Both grew from the astro-lab project at the University of Cologne.
2021
2 articles- They Told Me Not to Use Design Thinking. They Were Right. I treated a design workflow as though it were a complete research methodology. Grounded-theory techniques would have made my analysis of unexpected student responses more systematic. That is a postmortem of this thesis, not a universal verdict against design-based research.
- Mission to Mars: Teaching Air Pressure with a Smartphone and a Vacuum Pump You give students a vacuum pump, a bag of household materials, and a smartphone running phyphox. Their task: build a spaceship prototype that will survive the pressure difference between the crew compartment and space. A design-based inquiry experiment published in The Physics Teacher in 2022, and one of the more memorable experiments I have been part of running.
2020
3 articles- Please Stop Saying the Sun Is on Fire In September 2020 I gave a teacher training talk on stellar formation and the misconceptions students bring into class. The misconception list was long enough to be its own document. Here it is, with commentary. Includes: the Sun as a heat-planet, gravity that only works when things move, metals that always existed, and the obligatory complaint about quantum leaps.
- What Happens When You Film Student Teachers: ViLLA and the Case for Video in Teacher Education ViLLA is an online portal of real classroom videos built for teacher education at the University of Cologne. The idea sounds straightforward. Getting there required filming actual lessons, building infrastructure, evaluating a quasi-experimental intervention, and later scaling the project. Some notes on how that went.
- Hello World — What Is This Blog? An introduction to this blog: scientific ideas too lazy to submit, informal public review invited, with substantive corrections tracked in article changelogs and research notes.