{"pilot":true,"measured_on":"2026-10-08","list_prices_read_on":"2026-10-09","runs_per_question":1,"what_rule_right_means":"the answer states the decision rule of the node that covers the question, as graded by a model against keys written from the nodes; it is agreement with this product's rule, not an outside standard of correctness","arms":{"without":"the model alone, no skill, no node","free_layer":"the skill installed and told to use it: the free layer and the handles, no paid text","with_node":"the skill plus the paid node text for the question's decision"},"your_own_cost":"multiply tokens_per_answer by your model's list price: input_total counts every input token across the calls one answer took (most of it cache reads on the second and later calls), input_new the uncached part, output the tokens written; the measured cost_per_answer_usd used the prices in models.list_price_usd_per_mtok","will_change":"the comparison test (three runs per question, a second vendor, about sixty runs per arm) replaces these figures; measured_on changes when it does","how_to_read":"pick your model's class and the kind of question; the verdict is one of use (buy the node), apply (use the computed route, not the text), unproven (no lift established at this sample size), not_measured (no figure for your model), skip (outside the scope)","kinds":{"screen":{"label":"screens and thresholds","decisions":["risk/decide/hurdle","risk/decide/concentration","growth/decide/adjacency","growth/decide/pace","cash-flow/decide/owner-pay","value/decide/build-or-run-for-cash"]},"arithmetic":{"label":"arithmetic over periods, or an order of steps","decisions":["cash-flow/decide/runway","cash-flow/decide/cash-cycle","value/decide/reinvest-or-distribute","growth/decide/price"]}},"recommended_path":{"screen":{"path":"a small model with the node","what":"pin a small model for these questions and buy the node for each: the measured rule-right rate is the highest of any arm, at the lowest cost per rule-right answer","rule_right":{"small_model_with_node":"12 of 12","frontier_model_alone":"1 of 5","frontier_model_with_node":"4 of 5","across_all_kinds":{"small_model_with_node":"15 of 20","frontier_model_alone":"2 of 8","frontier_model_with_node":"5 of 8","note":"small model with the node against frontier with the node: no difference detected (p = 0.65)"}},"cost_per_rule_right_answer_usd":{"small_model_with_node":0.00472,"frontier_model_alone":0.43475,"frontier_model_with_node":0.18081,"across_all_kinds":{"small_model_with_node":0.00644,"frontier_model_alone":0.3352,"frontier_model_with_node":0.2412}},"the_delta":"per rule-right answer on screens and thresholds: $0.0047 for a small model with the node against $0.435 for a frontier model alone (92×) and $0.181 for a frontier model with the node (38×); across all kinds $0.0064 against $0.335 (52×) and $0.241 (37×)","if_you_cannot_choose_the_model":"on a frontier model already: the node's lift is unproven at this sample size (see the verdict); on a small model: use the node; on a model not measured: no figure — the small-model row is the nearest evidence, not a promise","quality":"quality here is rule-right: the answer states this product's rule for the question, graded by a model against keys written from the nodes — a pilot of one run per question, not a customer-satisfaction measure (a blind-preference arm is planned)"},"arithmetic":{"path":"the applied route","what":"POST the inputs to /apply/<decision>: the rule runs deterministically on your numbers for the node's price; any model can assemble the inputs; no model is needed to get the figures right","why":"on arithmetic over periods, or an order of steps the node's text did not lift the answers of either model (small model 3 of 8 with the node against 3 of 8 on the free layer; frontier 1 of 3 either way), so buying the text for a model to read is not the path; the computed verdict is","cost_per_rule_right_answer_usd_from_text":{"small_model_with_node":0.01333,"frontier_model_alone":0.23568,"frontier_model_with_node":0.48276},"applied_route":"the price of the decision's node per call plus your model's tokens to assemble the inputs; its answer rate was not part of the pilot (it computes, it does not read)","quality":"quality here is rule-right: the answer states this product's rule for the question, graded by a model against keys written from the nodes — a pilot of one run per question, not a customer-satisfaction measure (a blind-preference arm is planned)"}},"models":{"claude-haiku-5.5":{"class":"small","list_price_usd_per_mtok":{"input":0.1,"output":0.5}},"claude-opus-5.5":{"class":"frontier","list_price_usd_per_mtok":{"input":4,"output":20}}},"by_kind":{"screen":{"claude-haiku-5.5":{"n":12,"rule_right":{"without":2,"free_layer":7,"with_node":12},"cost_per_answer_usd":{"without":0.00234,"free_layer":0.00449,"with_node":0.00472},"cost_per_rule_right_answer_usd":{"without":0.01404,"free_layer":0.0077,"with_node":0.00472},"tokens_per_answer":{"without":{"input_total":18894,"input_new":7333,"output":1507},"free_layer":{"input_total":87758,"input_new":12791,"output":2368},"with_node":{"input_total":75373,"input_new":14107,"output":2579}},"api_calls_per_answer":{"without":1.17,"free_layer":4.33,"with_node":3.67}},"claude-opus-5.5":{"n":5,"rule_right":{"without":1,"free_layer":3,"with_node":4},"cost_per_answer_usd":{"without":0.08695,"free_layer":0.15456,"with_node":0.14465},"cost_per_rule_right_answer_usd":{"without":0.43475,"free_layer":0.2576,"with_node":0.18081},"tokens_per_answer":{"without":{"input_total":14974,"input_new":7543,"output":1256},"free_layer":{"input_total":80462,"input_new":12811,"output":1929},"with_node":{"input_total":76535,"input_new":12638,"output":1540}},"api_calls_per_answer":{"without":1,"free_layer":4.2,"with_node":4}}},"arithmetic":{"claude-haiku-5.5":{"n":8,"rule_right":{"without":1,"free_layer":3,"with_node":3},"cost_per_answer_usd":{"without":0.00249,"free_layer":0.00461,"with_node":0.005},"cost_per_rule_right_answer_usd":{"without":0.01992,"free_layer":0.01229,"with_node":0.01333},"tokens_per_answer":{"without":{"input_total":16197,"input_new":6516,"output":2185},"free_layer":{"input_total":67458,"input_new":12010,"output":2910},"with_node":{"input_total":67029,"input_new":13825,"output":3413}},"api_calls_per_answer":{"without":1,"free_layer":3.38,"with_node":3.25}},"claude-opus-5.5":{"n":3,"rule_right":{"without":1,"free_layer":1,"with_node":1},"cost_per_answer_usd":{"without":0.07856,"free_layer":0.13726,"with_node":0.16092},"cost_per_rule_right_answer_usd":{"without":0.23568,"free_layer":0.41178,"with_node":0.48276},"tokens_per_answer":{"without":{"input_total":15053,"input_new":5765,"output":1530},"free_layer":{"input_total":68385,"input_new":10816,"output":1962},"with_node":{"input_total":76730,"input_new":12674,"output":2337}},"api_calls_per_answer":{"without":1,"free_layer":3.67,"with_node":4}}}},"overall":{"claude-haiku-5.5":{"n":20,"rule_right":{"without":3,"free_layer":10,"with_node":15},"cost_per_answer_usd":{"without":0.0024,"free_layer":0.00454,"with_node":0.00483},"cost_per_rule_right_answer_usd":{"without":0.016,"free_layer":0.00908,"with_node":0.00644},"tokens_per_answer":{"without":{"input_total":17815,"input_new":7006,"output":1778},"free_layer":{"input_total":79638,"input_new":12479,"output":2585},"with_node":{"input_total":72035,"input_new":13994,"output":2912}},"api_calls_per_answer":{"without":1.1,"free_layer":3.95,"with_node":3.5}},"claude-opus-5.5":{"n":8,"rule_right":{"without":2,"free_layer":4,"with_node":5},"cost_per_answer_usd":{"without":0.0838,"free_layer":0.14807,"with_node":0.15075},"cost_per_rule_right_answer_usd":{"without":0.3352,"free_layer":0.29614,"with_node":0.2412},"tokens_per_answer":{"without":{"input_total":15004,"input_new":6877,"output":1359},"free_layer":{"input_total":75933,"input_new":12063,"output":1941},"with_node":{"input_total":76608,"input_new":12652,"output":1839}},"api_calls_per_answer":{"without":1,"free_layer":4,"with_node":4}}},"decisions":{"cash-flow/decide/cash-cycle":{"kind":"arithmetic","verdict_small_model":"apply","verdict_frontier_model":"apply","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=cash-flow/decide/cash-cycle"},"cash-flow/decide/owner-pay":{"kind":"screen","verdict_small_model":"use","verdict_frontier_model":"unproven","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=cash-flow/decide/owner-pay"},"cash-flow/decide/runway":{"kind":"arithmetic","verdict_small_model":"apply","verdict_frontier_model":"apply","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=cash-flow/decide/runway"},"growth/decide/adjacency":{"kind":"screen","verdict_small_model":"use","verdict_frontier_model":"unproven","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=growth/decide/adjacency"},"growth/decide/pace":{"kind":"screen","verdict_small_model":"use","verdict_frontier_model":"unproven","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=growth/decide/pace"},"growth/decide/price":{"kind":"arithmetic","verdict_small_model":"apply","verdict_frontier_model":"apply","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=growth/decide/price"},"risk/decide/concentration":{"kind":"screen","verdict_small_model":"use","verdict_frontier_model":"unproven","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=risk/decide/concentration"},"risk/decide/hurdle":{"kind":"screen","verdict_small_model":"use","verdict_frontier_model":"unproven","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=risk/decide/hurdle"},"value/decide/build-or-run-for-cash":{"kind":"screen","verdict_small_model":"use","verdict_frontier_model":"unproven","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=value/decide/build-or-run-for-cash"},"value/decide/reinvest-or-distribute":{"kind":"arithmetic","verdict_small_model":"apply","verdict_frontier_model":"apply","fit_url":"https://business-frameworks.matryoshka-paradigms.workers.dev/fit?decision=value/decide/reinvest-or-distribute"}},"not_for":["market data, prices or forecasts — the nodes carry none","businesses above about $50M in revenue, or questions outside value, cash flow, risk and growth","arithmetic over periods, or an order of steps answered from the text alone — the applied route computes them; the text does not","a frontier model already on a screen-or-threshold question — no lift was established at this sample size"],"source":"Work/tools/ab_measure (questions.json, results/summary.json); public kit without keys at github.com/matryoshka-paradigms/business-frameworks/measurement"}