facets · docs
‹ Index
DocumentLIVE

The translation apparatus, running — walkthrough for 2026-08-19

Status: LIVE — current as of 2026-08-18 (post #345 bench, milieu E=⟨O,M⟩, and the fresh-eyes polish pass — descriptions and units now authored on every element). Every number below is reproducible: cargo run -p bert-cli -- run assets/examples/translation-apparatus.sl --t 15.

0. One model, self-contained

The file is the whole base state. Structure, rates, the pool's starting level and its drain — all declared in SL (flows enter through their declared pass-ways, #226):

component "tRNA Pool" primitive Buffering stock tRNA initial 100 release 20
flow Nucleus -> "mRNA Entry Channel" : matter "mRNA transcript" amount 20 unit codon/s
flow Cytosol -> "GTPase-Associated Center" : energy "GTP" amount 40 unit GTP/s

The run card beside the diagram (and the Readouts rail) is where the knobs turn; a CSV is for data — measurements to validate against, a time-varying forcing a constant cannot say, an observed initial condition. Nothing about the base run needs one.

1. The structure receipt — his corrections, on the record

The 8/12 live demo drafted a ribosome from George's own description, and he located three errors. All three are in this model, as structure:

  • The polypeptide is a product, not a sink. Exit Tunnel → Chaperone carries "nascent polypeptide" out to its downstream consumer.
  • tRNA and GTP are inputs, not sources. GTP enters from the cytosol; tRNA is not an input at all at this boundary — it circulates: charged out of the pool, spent at the decoding site, recharged by the synthetase.
  • The ionic milieu genuinely interacts. Now said with his own newest construct — E = ⟨O, M⟩ from the lifecycle paper's introduction: milieu "Mg2+ and ionic milieu" value 1 unit mM. Ions have no point source and no port, so this is a condition that bathes, not a flow that plugs in. (The earlier flow encoding is the history: the kernel REFUSED that edge until the PTC was declared an interface — the same boundary check that fired on his call — and the milieu construct is the structurally honest resolution.)

2. The numbers, and where they come from

Steady-state elongation at 20 residues/s (17–21 aa/s prokaryotic). Energy enters twice — 2 GTP per residue at delivery/translocation, ATP→AMP (2 bonds) at charging — 4 bonds per residue, the textbook 4n−1. Initiation, termination and proofreading sit below this grain, and the model says so.

3. What building this model fixed in the engine

The honest-instruments section: quantifying the ribosome exposed three engine defects, each now fixed at the root with a separating test, or filed.

  1. The tap rule was substance-blind (#337). Any stock→sensor wire read as a non-draining level tap, so the pool minted matter — Pool' = 1.25·Pool + 45, verified in closed form. Now: a tap must be declared informational (bank-run's own convention — "read as a signal, not moved as stuff"); declared matter drains. The dt-invariance gate then caught the homeostat demo's own mis-declared sensed edge.
  2. The validation readout compared against the sender's total (#338). A multi-outwire source's whole emission was attributed to each flow — a fake "305% off" on a perfectly forced 20. Now each flow compares against its own wire's delivery; the forced flows read 0% off.
  3. Energy summed into output mass (#340). The synthetase made ~45 tRNA/s out of 20 matter + 20 ATP. Fig 3.17 rules it — processes transform substance "into high-quality versions of the same"; energy drives the work and leaves as waste heat — and the engine now says the same: 5 matter
    • 7 energy into a material Combining yields 5 out, 7 dissipated, never 12.

4. The run

--t 15 (one ~300-residue protein at 20 aa/s; re-measured 2026-08-18 on the current model):

  • Conserved, with every channel non-negative: emitted 1500 (+100 initial stock), sunk 37.5, stored 250, dissipated 1312.5, balance 0. The dissipated channel is large and honest — it is the GTP and ATP that drove the work, plus split losses. The 37.5 sunk IS the headline "polypeptide output · running total" on the run card.
  • The pool starts at its declared 100 and drifts +10/tick. The drift is a grain statement, not a defect: the decoding site's agency (0.5) and its even outwire split halve the tRNA return leg at this resolution. True stoichiometric assembly (output = the limiting input, not the sum) is a named future semantics, not something this model fakes with tuned numbers.
  • One caveat the ledger teaches: balance ≈ 0 certifies the ledger's own arithmetic, not conservation by itself — conservation is the non-negative channels plus the balance, which is what the #337/#340 fixes bought.

In the app: open Translation Apparatus → press ▶ Run in the strip under the diagram (the trace auto-plays once; ⏮ rewinds). The run card beside the diagram carries the sliders, the headline ("polypeptide output · running total") and the one chart; ⤢ opens the full-page Readouts. Click any processor for its mechanism: its description, its flows, and its actual per-tick rule as the engine computes it (the Combining line states the no-limiting-factor caveat in writing). To validate against data or force a perturbation (amino-acid starvation, seconds 8–10), attach a CSV in Data mode and bind columns — the pool can also take its t0 from a measured observation instead of the declared 100, and the two are the same kind of fact by construction.

5. The two walls — closed, and how

Until 2026-08-15 SL could not say a stock that starts full or a release rate — the archived reservoir and homeostat are the witnesses, and the pre-SL demos stayed unportable because of it. The #112 gate held until the transition functor was derived from the shipping stepper (F = Id over the circuit state; an initial stock is a choice of basepoint, a release rate a parameter of the map — neither changes F). Syntax then followed proved semantics: stock <unit> initial <n> and release <n>, with the emit refusal narrowed key by key, each narrowing carrying a separating instance. The tRNA pool is the first user. The refusals were evidence; the productions are the payoff.

6. The two senses of dynamics — his distinction, kept

What runs here is sense 1: configuration fixed, run Δt, measure. Sense 2 — the structure itself changing — is the paper's subject, and the tool does not pretend to it. The pool's level changes; the pool does not appear or dissolve. That seam is where the life-cycle formalism and this instrument will meet.