What We Learned
WHAT THE SYSTEM CAN LEARN
Which sources and repeated patterns are actually useful?
Evidence scores measure how complete the saved research is. They do not claim that a source or research area predicts markets.
How complete each publisher’s saved records are
A complete saved record is not a truth or forecasting score. It only says whether the required information was preserved.
Open every research area’s saved checks
Results by prediction type
Only individual saved results with the required proof count here. This breakdown is separate from the official overall score.
Repeated patterns with the most saved events
Different story links do not prove unique information. How much a source added, whether opposing evidence improved a decision and whether a market relationship helped a forecast remain unmeasured until their outcomes are linked.
How varied are today’s sources?
Publisher variety: 81.48 out of 100. CNA supplies 3 of 9 retained stories (33.33%), the largest publisher share in this edition.
Source locations · variety not measured
Languages · variety 81.48 out of 100
- en: 3 stories · 33.33%
- ar: 1 stories · 11.11%
- bn: 1 stories · 11.11%
- de: 1 stories · 11.11%
- es: 1 stories · 11.11%
- it: 1 stories · 11.11%
- th: 1 stories · 11.11%
Research areas · variety 79.01 out of 100
- Singapore News: 3 stories · 33.33%
- International News: 2 stories · 22.22%
- Banking & Credit: 1 stories · 11.11%
- Energy & Power: 1 stories · 11.11%
- Finance: 1 stories · 11.11%
- Manufacturing & Automation: 1 stories · 11.11%
Source locations describe the publisher or organisation, not necessarily the country of the reporter or the event. An unrecorded wire origin remains unknown.
Next improvement: do not score whether a source helped a prediction until enough released predictions have properly checked results.
The latest Bitcoin miss is an observed result from a declared experiment. It is not enough evidence to fit a more flattering confidence score. The current diagnostic is descriptive, and a small historical sample cannot establish a dependable trading edge.
The broader reading lesson is to keep unlike measurements apart: a task is not a whole job, a monthly production change is not a long-run trend, and a new commitment is not a completed benefit. These distinctions make a prediction testable before its outcome is known.
Graph response windows are measurements around dated events. They can show what prices did over a defined interval; they do not by themselves identify what caused the move. Missing windows remain marked as missing rather than becoming invented learning.
Original reporting and its limits
- CNA: Singapore minister distinguishes AI task replacement from the loss of whole jobs
These are a minister's assessment and historical graduate survey data, not a causal estimate of jobs lost to AI. Individual occupations and employers may differ. Retain the ministerial-forum report as an attributed assessment of task automation and graduate employment. The survey change is not a causal measurement of AI job losses. Accessible original article body and saved receipt; English summary limited to the observations stated here.
- tagesschau.de: German production falls as markets weigh oil and interest-rate risks
The production data describe July and the article is a 7 September update. One analyst notes a small three-month gain; market expectations are not central-bank decisions. Retain the current German market report. July industrial output remains historical data reported on 7September; oil and interest-rate commentary is contextual, not a realised forecast. Accessible original article body and saved receipt; English summary limited to the observations stated here.
See all 9 headlines and their sources →
What changed in the learning record
AAPL original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordBTC original August research review reconciled
The original 96-hour August research receipt records a missed direction and retains its eligible Coinbase candle.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordCL=F original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordCNY original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordETH original August research review reconciled
The original 96-hour August research receipt records a missed direction and retains its eligible Coinbase candle.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordEUR original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordGC=F original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordGOOGL original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordMETA original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordMSFT original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordNQ=F original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordNVDA original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordSOL original August research review reconciled
The original 96-hour August research receipt records a missed direction and retains its eligible Coinbase candle.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordTLT original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordUSD original August research review reconciled
The original August research receipt records an unverifiable result because its non-crypto endpoint method was not declared.
Next test: Use the list of issued decisions that do not already have a completed review and the exact original cohort method. Do not queue this completed receipt again.
This is an original completed review receipt reconciliation, not a new prediction result or historical founder score change.
See the underlying recordA missing L1 quote cannot be replaced with a candle
The saved price recovery Ethereum decision remains unverifiable because the required in-window L1 receipt was not captured.
Next test: Schedule observations under each issued contract; the new current publication terms declare a retrievable completed minute candle before issuance.
No old entry, method, outcome or founder total changes.
See the underlying recordBTC rule direction reversed before issuance
The same four-hour confirmation of twenty-hour direction changed from bearish at18:00Z to bullish at19:00Z. No prediction had been issued, so this is neither a miss nor a win. It is evidence that the short-horizon observation can change before an entry is frozen.
Next test: Freeze any actual new call only at its real issue time and retain the original future outcome rule. Compare later outcomes after they become due; do not change confidence automatically.
Observed preissue market change only. Original6/19 sign diagnostic, founder91 aggregate and all original outcome contracts remain unchanged. ETH and SOL also aligned positively but were not independently validated.
See the underlying recordBTC earlier fixed-rule diagnostic reviewed
Earlier observation at 2026-09-06T17:00:00.000Z: BEARISH. The 20-hour-spaced, sign-only diagnostic matched 6 of 19 directional samples. The later pre-issuance rule review at 2026-09-06T19:00:00.000Z supersedes this earlier Bitcoin direction.
Next test: Use the later pre-issuance review for the current decision. Only a real prospective issuance can create a future scored observation under its frozen terms.
Twenty-hour-spaced samples compare return signs only. They do not test the actual one-basis-point target, establish calibrated probability or causal evidence, or represent previously issued predictions.
See the underlying recordETH earlier fixed-rule diagnostic reviewed
Earlier observation at 2026-09-06T17:00:00.000Z: HELD, with no evaluated direction at that observation. The 20-hour-spaced, sign-only diagnostic matched 7 of 21 directional samples.
Next test: Use the later pre-issuance review for the current decision. Only a real prospective issuance can create a future scored observation under its frozen terms.
Twenty-hour-spaced samples compare return signs only. They do not test the actual one-basis-point target, establish calibrated probability or causal evidence, or represent previously issued predictions.
See the underlying recordSOL earlier fixed-rule diagnostic reviewed
Earlier observation at 2026-09-06T17:00:00.000Z: HELD, with no evaluated direction at that observation. The 20-hour-spaced, sign-only diagnostic matched 10 of 20 directional samples.
Next test: Use the later pre-issuance review for the current decision. Only a real prospective issuance can create a future scored observation under its frozen terms.
Twenty-hour-spaced samples compare return signs only. They do not test the actual one-basis-point target, establish calibrated probability or causal evidence, or represent previously issued predictions.
See the underlying recordCloudflare source evidence refreshed
Revenue rose from $639.8m in Q1 to $696.1m in Q2; annual growth increased from 34% to 36%. Current RPO growth increased from 34% to 35%. Q2 operating cash was $117.6m and free cash flow $56.4m. Q2 GAAP gross margin was 71.8%, below 74.9% a year earlier. GAAP operating loss widened to $205.7m from $67.3m. Cash and securities of $4,162.8m are a balance-sheet fact, not proof of recession survival or fair valuation.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordRocket Lab source evidence refreshed
Q1 revenue was $200.3m and Q2 $234.1m; backlog rose from about $2.2bn to $2.36bn. Q2 GAAP operating loss narrowed to $57.514m from $59.639m a year earlier. First-half operating cash use increased to $134.407m from $77.467m. Financing supplied $1,523.403m. Ordinary shares outstanding rose from 543.575m at December 31 to 598.180m at June 30. Large cash balances therefore do not establish self-funding growth. Neutron’s planned launch is not completed execution.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordTempus AI source evidence refreshed
Q2 revenue was $382.5m, up 22%. The issuer reports oncology volume growth of 31%, versus 28% last quarter, and MRD tests of 9,000 versus 6,500. It reports FDA approval of tumor-only xT CDx and a delivered AstraZeneca model. Net income of $5.6m included $98.5m of unrealized securities gains. GAAP operating loss was $75.913m; first-half operating cash use was $80.802m. Cash and securities were $820.7m, while convertible senior notes net were $1,172.781m. The Personalis transaction was an agreement, not proof of a completed acquisition.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordSolana source evidence refreshed
The official live status response reported all systems operational at the captured retrieval bracket. A point-in-time operational status does not measure long-run reliability, security, users, fees, token-holder value capture, concentration or valuation.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordOriginal Tamagotchi Sakura ordinary retail comparison
Counterexample: strong product similarities and nostalgia can coexist with modest later sale prices. This is a useful control in the watchlist, not an appreciation call.
Next test: Recheck this exact retail offer next edition; seek original 1997 price evidence and more condition-matched completed sales, preserving working, defective and sealed examples separately.
Sakura is a later shell, not an original 1997 unit. Reproduction of the old experience provides substitutes. Battery condition, functionality, packaging and authentication matter; eBay attribution is seller supplied. Original MSRP is missing, so neither a historical return nor future target is calculated.
See the underlying recordMedieval Horse Knight Castle ordinary retail comparison
A research candidate because it closely matches the historical product category. The available evidence supports a comparison, not a prediction of appreciation.
Next test: Recheck retail price and availability in the next edition; seek original 6080 retail pricing and several comparable used/complete sales, plus 31168 production and resale evidence.
Different design and generation; unknown production totals and original cost baseline. Preserved contents, packaging and condition materially affect the cited sale. Modern substitutes and future reissues may limit demand. Storage, fees, shipping and inflation prevent treating gross auction prices as profit.
See the underlying recordONCE AGAIN ordinary retail comparison
An incomplete-evidence control: the old model is identifiable and a listing completed, but its sale amount is undisclosed. Familiar traits cannot fill that evidence gap or justify a future price claim.
Next test: Recheck current stock and price next edition; obtain a dated primary GB101 catalog and authenticated, condition-matched completed sales before strengthening this comparison.
Different dial, reference and strap materials; continuous substitutes may limit demand. Seller dating and authenticity require further verification. Missing original retail price, transaction costs and inflation prevent a return calculation. Current Singapore retail and historical US-dollar prices are not a direct performance comparison.
See the underlying recordBTC original-rule outcome recorded
Append-only official observation: MISS. Original issued direction and terms are unchanged.
Next test: Review failure conditions using original terms. One outcome does not establish a calibrated forecasting edge.
Separate experimental cohort; this does not change the founder aggregate or legacy record membership.
BTC earlier fixed-rule diagnostic reviewed
Earlier observation at 2026-09-07T16:00:00.000Z: BEARISH. The 20-hour-spaced, sign-only diagnostic matched 11 of 19 directional samples.
Next test: Use the later pre-issuance review for the current decision. Only a real prospective issuance can create a future scored observation under its frozen terms.
Twenty-hour-spaced samples compare return signs only. They do not test the actual one-basis-point target, establish calibrated probability or causal evidence, or represent previously issued predictions.
See the underlying recordETH earlier fixed-rule diagnostic reviewed
Earlier observation at 2026-09-07T16:00:00.000Z: BEARISH. The 20-hour-spaced, sign-only diagnostic matched 11 of 19 directional samples.
Next test: Use the later pre-issuance review for the current decision. Only a real prospective issuance can create a future scored observation under its frozen terms.
Twenty-hour-spaced samples compare return signs only. They do not test the actual one-basis-point target, establish calibrated probability or causal evidence, or represent previously issued predictions.
See the underlying recordSOL earlier fixed-rule diagnostic reviewed
Earlier observation at 2026-09-07T16:00:00.000Z: BEARISH. The 20-hour-spaced, sign-only diagnostic matched 13 of 20 directional samples.
Next test: Use the later pre-issuance review for the current decision. Only a real prospective issuance can create a future scored observation under its frozen terms.
Twenty-hour-spaced samples compare return signs only. They do not test the actual one-basis-point target, establish calibrated probability or causal evidence, or represent previously issued predictions.
See the underlying recordCloudflare source evidence refreshed
Cloudflare announced invitation-only Vulnerability Discovery and Remediation with OpenAI Daybreak on September 3. This adds a concrete product distribution step to the rechecked Q1/Q2 demand evidence. Revenue rose from $639.8m in Q1 to $696.1m in Q2; annual growth increased from 34% to 36%. Current RPO growth increased from 34% to 35%. Q2 operating cash was $117.6m and free cash flow $56.4m. Early access for selected Enterprise customers does not demonstrate adoption, incremental revenue or superior margins. The original quarterly margin and loss concerns remain.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordRocket Lab source evidence refreshed
The current issuer index reports its 94th Electron mission succeeded on September 2. The August 10 Q2 release remains the latest financial release identified in this check. Q1 revenue was $200.3m and Q2 $234.1m; backlog rose from about $2.2bn to $2.36bn. Q2 GAAP operating loss narrowed to $57.514m from $59.639m a year earlier. A successful mission count is execution evidence, not proof of Neutron delivery, sustainable cash generation or a fair entry valuation.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordTempus AI source evidence refreshed
The August 24 issuer announcement adds FDA clearance for ECG-PH, its third cleared cardiovascular device, to the previously checked Q2 evidence. Q2 revenue was $382.5m, up 22%. The issuer reports oncology volume growth of 31%, versus 28% last quarter, and MRD tests of 9,000 versus 6,500. It reports FDA approval of tumor-only xT CDx and a delivered AstraZeneca model. The product is not a stand-alone pulmonary-hypertension diagnostic. Clearance alone does not establish adoption, reimbursement, clinical outcomes or cash profitability.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordSolana source evidence refreshed
The September 4 foundation roundup reports more than $4bn of real-world assets and a record 216 million non-vote transactions in one day in August. This is new retained adoption evidence; the fresh status response is separately operational. These are foundation-reported ecosystem measures, not independently reconciled fees or holder returns. Addresses and transactions are not equivalent to paying users, and infrastructure adoption does not establish SOL valuation.
Next test: Review remaining subtraits and valuation evidence before a complete score or ranking change.
Current partial evidence review; historical whole-rubric score remains separately dated.
See the underlying recordOriginal Tamagotchi Sakura ordinary retail comparison
Counterexample: strong product similarities and nostalgia can coexist with modest later sale prices. This is a useful control in the watchlist, not an appreciation call.
Next test: At the next edition, recheck this exact retail SKU, market, price and availability. Revisit the original historical comparison only when independently verified new sale evidence arrives; do not infer a return from undisclosed Best Offer settlement.
Sakura is a later shell, not an original 1997 unit. Reproduction of the old experience provides substitutes. Battery condition, functionality, packaging and authentication matter; eBay attribution is seller supplied. Original MSRP is missing, so neither a historical return nor future target is calculated.
See the underlying recordMedieval Horse Knight Castle ordinary retail comparison
A research candidate because it closely matches the historical product category. The available evidence supports a comparison, not a prediction of appreciation.
Next test: At the next edition, recheck this exact retail SKU, market, price and availability. Revisit the original historical comparison only when independently verified new sale evidence arrives; do not infer a return from undisclosed Best Offer settlement.
Different design and generation; unknown production totals and original cost baseline. Preserved contents, packaging and condition materially affect the cited sale. Modern substitutes and future reissues may limit demand. Storage, fees, shipping and inflation prevent treating gross auction prices as profit.
See the underlying recordONCE AGAIN ordinary retail comparison
An incomplete-evidence control: the old model is identifiable and a listing completed, but its sale amount is undisclosed. Familiar traits cannot fill that evidence gap or justify a future price claim.
Next test: At the next edition, recheck this exact retail SKU, market, price and availability. Revisit the original historical comparison only when independently verified new sale evidence arrives; do not infer a return from undisclosed Best Offer settlement.
Different dial, reference and strap materials; continuous substitutes may limit demand. Seller dating and authenticity require further verification. Missing original retail price, transaction costs and inflation prevent a return calculation. Current Singapore retail and historical US-dollar prices are not a direct performance comparison.
See the underlying recordOpen the complete outcome and learning record