Corrections and updates.

What was clarifiedThe methodology has been clarified to define a priced-provider observation issue and to state that Coverage Score and the Compute Opacity Index are not calculated when the complete panel is not adjudicated. Each priced-provider observation issue also discloses the most recent sealed full sweep and the date the next full re-verification is due. The clarification is effective with Issue 009.

What was wrongA source review found several launch-era statements that were too broad, mixed unlike time horizons, or used numerical comparisons without a consistent denominator. The four-trillion-dollar essay also combined calendar-year capex guidance with Oracle fiscal-year guidance and described total corporate capex as if every dollar were AI-specific.

What is rightThe affected essays now distinguish corporate capex from AI-specific investment, keep Oracle fiscal-year figures separate from calendar-year totals, use a single comparison basis, and label analysis as analysis. The founding essay now uses source-backed Lambda configuration examples rather than an unrecoverable seven-source claim. The Issue 004 essay retains its sealed issue measurements while tightening historical evidence framing.

How the error occurredEarly editorial copy combined launch research, later corporate guidance, and evolving product plans before the publication had a mature source-map and horizon-control process. The August 11 review applied the current sourcing, distinction, and correction standards to the historical pages.

What was wrongRateCheck rounded a $474.50 monthly benchmark gap to $474 because of display precision. Several institutional pages also used wording broader than the governing record, including an unsupported private-contract discount range, an overbroad permanent provider-funding statement, sponsorship copy tied to audience maturity rather than the issue-sequence firewall, and homepage positioning that stated an unsupported publication-of-record and largest-in-modern-history superlative as fact.

What is rightRateCheck now displays monthly dollar gaps to cents. SVCI private-pricing language no longer asserts an unsourced discount range. About, Press, and Sponsorships now describe the published editorial-commercial firewall and current Issue 010 and Issue 013 sponsorship gates accurately. Homepage positioning now describes Sorso View as an independent publication and frames its ambition without unsupported superlatives.

How the error occurredThese defects came from older launch copy and a JavaScript floating-point/display interaction that survived the Issue 007 deployment sweep. They were found during post-deployment source and live-page QA.

What changedThe former single-track cadence, under which a skipped weekly cycle could force a fresh twenty-seven-provider sweep, is superseded by A1-R1. The SVCI now uses a monthly full-panel Track 1 sweep and fresh priced-provider Track 2 captures on weekly or biweekly issue cycles.

How Track 2 worksPriced providers are observed fresh. Excluded dispositions carry from the most recent sealed full sweep and are explicitly disclosed with the full-sweep date and next re-verification due date. Skipped Track 2 slots no longer force a full sweep.

Issue 007 applicationIssue 007 freshly observed fifteen priced providers on August 10, 2026. Twelve excluded dispositions carry from the July 21 Issue 006 full sweep. The next full sweep is due on or about August 21.

What was wrongThe Issue 005 release-candidate metadata reported the Hyperscaler Premium as 2.9664x. That figure divided the Tier 1 median by the Tier 2 median.

What is rightThe locked formula divides the Tier 1 median by the pooled non-hyperscaler median. Using the sealed Issue 005 inputs produces 2.9498x.

How the error occurredA derivative metadata field used the Tier 2 median as the denominator instead of the pooled non-hyperscaler median. The Issue 005 page, calculation report, capture ledger, and data record now carry the corrected figure.

What was wrongThe Issue 004 and initial Issue 006 site files described Price Floor as the mean of four Tier 2 rates. That wording was not the locked methodology.

What is rightPrice Floor is the median of the lowest K headline observations, where K is the greater of one and ceil(N times 0.25), across all headline-included rates. Issue 004 remains $2.82 at public precision. Issue 005 is $2.82. Issue 006 is $2.97, not $3.02.

How the error occurredA later packaging pass substituted an operational Tier 2 mean for the formula locked in methodology v1.1. The site, current calculation report, and methodology page now use the locked formula.

What was clarifiedThe site previously used the phrases "three cadences," "quarterly research," and "four reports per year plus monthly research briefs" without clearly separating the weekly SVCI, the not-yet-active Sorso View Brief, periodic Essays, and the Reports program. The current site then stated that all three SVCI tiers were reviewed in one weekly issue, the Sorso View Brief is event-driven and not yet active, Essays have no fixed quota, and the Reports page describes a rolling production plan with separate monthly Reports Research Notes. No provider rate, historical issue result, methodology formula, or index value changed.

Update, August 25, 2026The cadence described in this entry was superseded on August 11, 2026 by the SVCI Two-Track Capture Cadence entry above. The full sweep referenced as due on or about August 21 was completed in Issue 008, published August 16, 2026.

What was wrong The original evidence block listed 2026 capital expenditure figures based on late-January and February 2026 analyst estimates:

• Microsoft listed at approximately $120B
• Alphabet at $175 to $185B
• Meta at $115 to $135B
• Oracle at approximately $30B
• Top 5 hyperscaler total at approximately $602 to $700B

What is right The May 3 correction properly updated the company guidance then available for Microsoft, Alphabet, and Meta. The August 11 source-integrity review refined the treatment further: Oracle's approximately $50B figure is FY2026 and is now kept separate from calendar-year totals; the mixed-horizon top-five aggregate, US-GDP comparison, and unsupported AI-specific subtotal were removed. The article now links directly to primary company sources and labels total corporate capex separately from AI-specific investment.

How the error occurred The essay was finalized using analyst-consensus estimates current as of late January and February 2026, before the April 29 earnings cycle. The April 29 revisions had not been incorporated when the essay went to publication on April 30. The editorial process did not include a final pre-publication check against the most recent earnings disclosures. We have updated the editorial workflow to require an explicit pre-publication verification against the latest earnings cycle for any piece referencing recent corporate guidance.