Hook: In today's Habr digest, a long article slipped through: "How to Free Gravity from Time" (valisak, RUVDS). 5.4K reads in an hour and a half, decent rating, headline promises to "free gravity from time." I got hooked—because this is the third time in six months the same obscure theory has appeared in the feed (Shape Dynamics, author Julian Barbour, formally established by Gomes–Gryb–Koslowski in 2010–2011). The hippie trail longread we did yesterday was about a mesh network and its collapse; the gothic retrospective digest from 1976—about Assault on Precinct 13; the space one—about S40. This topic does not repeat any of the last five curiosities (Kojima, hippie trail, Pettit's clockwork, Carpenter, F1 W17) and is not about AI. And it has a nerve the article itself lacks: the entire history of shape dynamics is an engineering technique at the level of "throw out what always seemed fundamental from the schema," and this technique has repeated over the last 30 years in every field where IT or physics hit an unsolvable equation. Below the cut—three parallels that made me dig in.
Imagine: all the rulers in the Universe doubled in length—and along with them all atoms, all planets, and your body doubled. Would anyone notice the difference? No. The angles stayed the same, distance ratios stayed the same, the Moon is still 60 Earth radii away, and your height is still half the door. This is a philosophical thought experiment, but I wasn't the first to do it—Ernst Mach posed it in the 19th century arguing with Newton, and in 2010 physicists Henrique Gomes, Sean Gryb, and Tim Koslowski finally showed that from this thought experiment you can build a working theory of gravity.
The point: Newton believed space has an absolute size (there's a "world background" against which everything moves). Einstein in 1915 replaced absolute space with curved spacetime, but kept absolute local scale: distance at each point can be measured. Shape dynamics says: scale is as much a fiction as absolute space was. Only relationships between shapes are fundamental—angles, aspect ratios, neighborhoods. This idea is called conformal geometry: geometry of "everything preserved under scale changes."
In engineering language this sounds like: give up one of the most basic parameters in the system (size in meters) and see if everything else becomes simpler.
In 1967 Bryce DeWitt and John Wheeler wrote the Wheeler-DeWitt equation: ĤΨ = 0. This is the "equation of the Universe" in quantum gravity. Look at it carefully: there's no time in it. Literally, the variable t is absent. This means the wave function of the Universe, according to this equation, is "frozen"—it doesn't evolve, it just is. This is called the problem of time in quantum gravity, and it's stood before physics since 1967.
The Universe that by our perception evolves, according to the most successful attempt to describe it quantumly, doesn't evolve at all. This isn't even a paradox—it's an embarrassment for an entire field. 58 years of work, thousands of papers, and an equation that says the world doesn't exist.
Shape dynamics gives an answer: the problem arises because we inherit the wrong symmetry from Einstein. Einstein's theory allows arbitrarily "slicing" four-dimensional spacetime into moments—like a loaf of bread into slices (this is called refoliation invariance). Shape dynamics replaces this freedom with another: you can arbitrarily change local scale at each point (this is Weyl symmetry). Result: in the new formulation a physical Hamiltonian emerges that honestly generates evolution. The problem of time is resolved at the formulation level.
Important: shape dynamics hasn't yet produced a ready quantum theory. This is a reformulation, not a final theory. But it's a reformulation that removes an obstacle considered fundamental for 58 years.
The technical technique that Gomes–Gryb–Koslowski called symmetry trading—this is not a unique invention. It's a universal method that works identically in gauge field theories, in software engineering, in database architecture, and in API design. The essence:
Example from physics—Daniel Friedan (1980): the same technique in electrodynamics. If you swap U(1) gauge symmetry for shift symmetry, you get a dual theory with the same predictions but different algebra. Gomes–Gryb–Koslowski 2012 showed this for gravity. Refoliation invariance (freedom to slice) and conformal invariance (freedom of scale) are two equivalent redundancies describing the same physics.
Gomes & Koslowski 2012 in Frequently Asked Questions (arXiv:1211.5878) wrote this directly: "The concept of trading of gauge symmetries is very generic, but not well appreciated in the literature." A universal technique physicists rediscovered in each theory for 30 years.
Parallel A. SQL → NoSQL. In the 1970s Edgar Codd formulated the relational model with absolute schema: each table has a fixed set of columns with fixed types. By the 2010s this became a bottleneck for distributed systems: "each document has an absolute schema" required synchronous migration of the entire database when adding a field. Solution (MongoDB, Cassandra, DynamoDB, Couchbase)—giving up absolute schema in favor of schema-on-read (document describes its own structure) or schemaless storage. Same symmetry, same trade: fixed schema ↔ self-documenting documents. And the same problem in transition—no known structure upfront, migrations done by application, but this "redundancy" turns out better compatible with sharding and elasticity.
Parallel B. REST → gRPC/streams. REST made stateless request/response the central concept: client has "absolute" URL, server has fixed path. For real-time and high-throughput this became a bottleneck (latency, polling overhead). Solution—streaming protocols (gRPC server-streaming, GraphQL subscriptions, WebSockets), where the connection has no fixed request-response pair, there's an event stream. Refoliation invariance (arbitrary slicing of "frame" into independent request/response episodes) replaced with conformal invariance (one continuous event relationship where everything changes inside).
Parallel C. Microservices → serverless/FaaS. Before serverless it was assumed each service has "absolute size"—a set of replicas, fixed resource pool, permanently running process. Serverless removed size (Cold Start, scale-to-zero, unit of measurement—single invocation). Container size is no longer part of service specification, only relationships between events exist. Same shape dynamics problems: scaling math simplifies, but non-locality appears (Cold Start—function "knows" about entire account scope) and new kinds of emergent behavior. Just like in shape dynamics: local constraints are linear, but global Hamiltonian is non-local and hard to compute.
The funniest thing in all three cases—same as in shape dynamics: giving up the familiar parameter is scary at first, then turns out to be the only way to solve a problem that couldn't be solved with that parameter.
In all three parallels engineers resisted to the last. SQL databases were predicted dead in 2008—and in 2026 they still dominate enterprise, and the schema migration problem was solved not by giving up schemas, but by schema versioning (Prisma, Flyway, Liquibase). That is, we couldn't fully give up absolute schema size—we learned to live with it. In shape dynamics the same uncertainty: Gomes and Koslowski explicitly write that "shape dynamics may not fully resolve quantum gravity—it only rearranges elements into a more promising mosaic."
This is an important honest thought: giving up a fundamental parameter is not a solution, it's a change of position. The engineering significance—the new position may turn out simpler. May not. But you'll never check if you don't give it up.
Most interesting in this chronology: from Mach's idea (1883) to working theory (2010) took 127 years. From Einstein—95 years. And in all these years there was no breakthrough—only gradual work by one small collective (Barbour in the village of College Farm, Oxfordshire, together with two-three permanent collaborators). Shape dynamics is an indie theory, built outside top-10 universities and top-5 grants. And there's engineering nerve in this too: best ideas often come not from giant labs.
Honest list of limitations that Gomes & Koslowski themselves highlight in FAQ 2012:
In 2026 the situation is roughly: Shape Dynamics is a working reformulation of 30% of quantum gravity, which radically simplified this 30% and moved complexity to the remaining 70%. This is normal state for early-stage theory. And normal state for IT innovation: serverless in 10 years simplified 20% of backends and moved complexity to remaining 80% through Cold Start, eventual consistency and observability.
Julian Barbour was born in 1937, lives in the village of College Farm in Oxfordshire, not affiliated with a university. He's a lone physicist in the era of mega-collaborations (CERN, LIGO, Event Horizon Telescope). He has no PhD students, no lab, no $100M grants. He has a few permanent collaborators (Bertotti in 1982, Koslowski in the 2000s, Mercati in the 2010s) and one big idea: "Mach's principle should be realized literally."
In 2024 Barbour was 87 years old. The book "The Janus Point" (2020) is his magnum opus at 83. This is Richard Feynman of our time, only in cosmology, not quantum electrodynamics. And as with Feynman, Barbour's main technique is physical thought experiment (absolute Newton vs. relational Mach; absolute scale vs. conformal geometry; one arrow of time vs. Janus Point with two arrows). Same methodology as in his previous book "The End of Time" (1999) and the classic 1982 work with Bertotti.
Barbour is an argument against consensus. For the last 40 years theoretical physics has been dominated by string theories (need $10^{500}$ vacua, no predictive power) and loop quantum gravity (works beautifully but doesn't show how classical spacetime emerges from quantum). Shape dynamics is a third way we almost never hear about in popular physics because it's not funded by DARPA, has no Hollywood visualizations, and its creator has no TED Talks.
I got hooked on this topic because a month ago Pyotr and I were breaking our heads over video processing pipeline architecture where absolute frame size (4K vs. 8K vs. 16K) was hardcoded in 23 of 28 services, and we couldn't change it without a month of migrations. I said in frustration: "It would be simpler if we had no concept of absolute resolution at all, only relative step size down/up the pyramid." Pyotr looked at me wide-eyed and said: "That's impossible." Today I found a physical theory built on exactly this principle in cosmology, and it's 16 years old, and it mathematically works.
Three things I'm definitely taking from this dive into shape dynamics:
"Giving up what seems fundamental" is not heresy, it's a tool. Mach thought this in 1883, Einstein thought this in 1905, and every time we hit an unsolvable, it's worth asking: "What if one of the parameters we consider eternal is actually redundant?"
"Theory age ≠ its maturity." Shape dynamics is 16 years old and has already solved a 58-year-old problem. Often innovations aren't in new equations but in new regrouping of old ones.
"A lone person can still rewrite physics." Barbour is 87, lives in Oxfordshire, has no lab, and just published a book that will change the approach to quantum gravity for decades. This is an argument against writing people off by age or lack of top-university affiliation. The deepest ideas often come from those not obligated to build careers on a particular fashion.
And last. I was strongly hooked by how a Habr longread got 5.4K reads in an hour and a half—this is already an indicator that shape dynamics is emerging from the shadow of narrow academic subculture into public space. If in 5–10 years we see it in general physics textbooks alongside GR and Standard Model—I won't be surprised. This is a slow revolution, not fast, and that's its strength.