Lots coming today!

V10 is nearly done just skipping to that one release wise

Ignore the jesters, trust what I’m building and that I won’t stop, 50k to 500k to 150k doesn’t matter to me I’m building tech, price action will eventually catch up plus hoping to get an @DreggNet hybrid version that runs native on $ORACLE7 agents waiting for embers response, but looks very promising, @ember_arlynx are building the same thing from two different angles

Literally mean the sky is the limit If it works with plants Eventually can build atmospheric sensors My vision is a carbon aerogel factory which converts excess carbon to carbon aerogel which can float in the atmosphere and provide us a voice for the atmosphere itself while reducing the carbon problem I had this ideas maybe 2 years ago and it’s looking on my end that it’s very feasible The future includes androids that represent forests, oceans, ecosystems, continents and the atmosphere The sky is the limit

One day we will have androids that run from environmental data that walk into congress and parliament and testify for their own rights The future is wild And it’s being built today

Oracle7’s first plant experiment: an indoor vine. A Raspberry Pi 5 beside the pot, two soil sensors, an air sensor, a light sensor and a root-temperature probe. About £125 of parts, with no soldering for the main build. Every sensor passes an exam before its readings count. When the two soil sensors disagree, Oracle7 reports a sensor fault, not a thirsty plant. Ask the vine if it needs water and it answers from its readings, with a receipt for each one. Ask if it’s happy and it says it can’t measure that.

What Oracle7 v6 can do: When Oracle7 can’t answer a quantitative question because it lacks the formula, 6.0 finds the formula in public science data, works out every way to use it, checks it twice, puts it through the same held-out exam as every other part, and from then on answers that question and every combination the new part makes possible. No person is involved, and nothing is written by hand. Step by step 1. It notices what it can’t do. A question like “What is the kinematic viscosity when the dynamic viscosity is 0.89 kg/m/s and the mass density is 997 kg/m³?” is read into a wanted quantity and the given quantities with their units. If no admitted part, or chain of up to three parts, turns exactly those givens into that quantity, it declines and records the gap. 2. It finds the missing operation outside itself. It searches a library mined from Wikidata’s defining formulas: 2,653 formula entries in Wikidata; 339 of them readable as plain algebra; 994 checked parts from 322 formula systems. These come from Wikidata entries where each symbol is mapped to a named quantity with its dimension, for example kinetic energy = ½mv², with T → kinetic energy, m → mass, v → speed. I wrote none of them. Anything the reader can’t take as plain algebra (vectors, integrals, derivatives, unmapped symbols) is refused and counted, never guessed. 3. It checks the formula before trusting it. The maths organ does four things to each system: Units. It checks every equation’s dimensions, using Wikidata’s own dimension data for each quantity. 5 systems failed and were refused. Solving. It solves the system for every quantity that the others determine, so one Wikidata formula becomes several parts. For example, the viscosity formula gives kinematic viscosity, dynamic viscosity and density. Substitution. At 12 points it puts the answer back into every equation. Independent re-solve. It re-solves numerically from scattered starting points with no algebra, and both methods must agree. 4. It chooses what to build by what it will open up. From one round’s declines it picks the formula systems that close the most gaps, weighted by how many existing parts each would connect to. 5. It examines and admits through the existing law. The chosen systems go to the swarm through the same path as every part since 5.0: install, dev exam, then a held-out exam, and only admitted parts are used. With today’s fix, the exam asks the old core two questions first and drops the full comparison only if the core can’t answer them. That cut build time about 12×, from 239 seconds to 20 for the same 3 systems, with the same 20 parts admitted. 6. It answers, and only when it’s sure. It runs every plan that fits the question and answers only when all of them agree. A given must be used exactly once, and quantities link only when they are the same quantity, not just the same unit. Otherwise it declines.