Friday September 4 settled corpus, published September 6 for September 8. Monday September 7 has no bar. Clean read-only analyst owns every admission; main owns acquisitions, code verification and persistence. The screen nominates; company evidence admits.

Coverage and price grid

467 = original 464 plus WULF, GLXY and CBRS omitted by the first union. All ten neoclouds are now included. GOOD 23; EXTENDED/unconfirmed 313; BROKEN 102; excluded 29. The original 464 denominator had 23 / 312 / 101 / 28. CBRS has insufficient long history; a missing SMA200 excludes it rather than making it broken. Weekly-only, stale, held and nonpositive-volume inputs each retain their exclusion reason.

classify.py reads the frozen JSON inputs and calculates the grid. GOOD requires above SMA200, current RSI30–45, at least three down days with lighter mean volume than the prior twenty-session baseline, and a sustained recent pop. BROKEN is below SMA200 or a volume-confirmed heavy breakdown. EXTENDED also includes unconfirmed entries; it does not mean every row is parabolic. This diagnostic is local to the campaign and changes no engine gate.

Company admission and entry conditions

Every watch below has a numeric price alert, not an automatic BUY switch. The five HIGH rows pass the narrow company thesis; their jagged cohort still blocks BUY. The ten LOW rows additionally require fresh company evidence. They retain named evidence gaps rather than receiving ten unnecessary deep dives. Only MU was commissioned as a dive, within the cap of eight.

Review every BUY/WAIT on September 11.

Outside this capital window

The GOOD screen-only nominees ABBNY, ADI, AIR, DAL, GE and IRM are AVOID for this campaign because no company thesis was admitted. The remaining universe retains its diagnostic row, not an invented recommendation.

Worked application

Published September 6 for the Tuesday September 8 window. Monday September 7 is the Labor Day holiday. Memory and electricity remain the two operating primaries; physical-AI components return to the bench pending attributable component economics. This campaign applies the book-blind, levels-only override in Campaigns.

The complete screen covers 467 distinct names: the original 464 plus WULF, GLXY and CBRS, restoring all ten declared neoclouds. It nominates 23 GOOD, 313 EXTENDED or unconfirmed and 102 BROKEN entries; 29 are excluded for missing, stale, unsound or insufficient canonical inputs. These are price classifications, not company convictions. The diagnostic compares down-day volume during the most recent completed twenty-session window with the preceding twenty-session baseline and requires a sustained recent pop. Every exclusion and formula is saved in the Names receipts.

GE Vernova is the supported primary name among the strict GOOD nominees, but its jagged cohort keeps it at WAIT: reclaim 971.98, then hold a flag; 886.12 is the longer-trend invalidation reference. SanDisk also waits despite current memory demand evidence. Constellation's contracted future power supports the electricity thesis, while its price rehabilitation still needs a held flag. Nebius and IREN have company-relevant AI-cloud delivery evidence, but both remain jagged timing waits; the eight other neoclouds are excluded from this capital window until their distinct operating and financing questions are resolved.

A price screen cannot convert the remaining older basket names into admitted thesis purchases. TSM, SOXX, AMAT, NVDA, WDC, GLW, COHR, MRVL, LRCX and VRT remain evidence-pending watches with explicit levels and a September 11 review. Their prices can trigger review, not automatic admission. Together with IREN, they supersede all eleven existing weekly calls. Historical source calls remain unchanged for grading; predecessor targets are closed and linked to one new campaign target per name. There is no scheduled accumulation.

Micron is the one BUY: its historical bottom, pop, contracting flag and September 4 breakout support a manually adjudicated GOOD entry. The strict screen remains EXTENDED and the August 10 RSI reset was 46.30, just above the approximate 30–45 guide. This exception is explicit. BUY permission requires holding the cleared flag ceiling at 989.96; a lost ceiling returns it to WAIT. Pattern failure is 887.61; the distinct long-trend invalidation is SMA200 at 606.48. The September 6 company dive supplies current admission, with June filing dates preserved. No normalized fair-value claim is made.

TSM's August 28–September 3 flag was 407.82–429.12, with mean volume 0.7324× the preceding twenty-session mean. Friday's 428.91 close remained below that flag high and the August 27 pop high 429.64. The canonical 461.47 breakout is unhit. Its historical RSI low 45.80 was near the guide; today's 56.77 is not the reset value. Company evidence is also pending.

Full decisions, volume receipts and exclusions: September 6 Names read. Price authority: research/market-engine/data/summaries/*.json, September 4 settled, with canonical-bar extracts in the Names receipts. The manually reviewed diagram does not change engine rules or the strict diagnostic output.

Evidence, acquisition and limits

• Reproduce the diagnostic: python3 research/investigations/2026-09-06-campaign-01-names-receipts/classify.py --inputs research/investigations/2026-09-06-campaign-01-names-receipts.
• The frozen receipt folder carries all 467 classifications, exact canonical bars, original summary rows, historical RSI, inputs, hashes, analyst decision JSON and target identities. No account inputs are used.
• MU fresh pulls: Massive financials, EDGAR facts, show/leading, balanced deep-dive, fresh forced news, both bare-ticker and qualified social, and the FY2025 10-K. Latest June 25 10-Q is on disk and verified. Eight discrete quarters are reconciled, with missing Q4 EPS explicitly held. News force returned 0 new /5 existing; it was a completed search, not an empty corpus. Social is thin and older material, not evidence of positioning consensus. Selective transcript link redirected; no snippet was promoted to a retrieved transcript.
• Price/history sources: research/market-engine/data/summaries/*.json; canonical readDaily extracts in 2026-09-06-campaign-01-names-receipts/entry-bars.json and entry-supplement.json. No direct OHLC-file reader was added.
• Company/event sources: Theses read and MU dive; issuer dates remain original. Phase2 evidence-gap commissions remain needs-evidence.

Sources

• Micron June 25 Form10-Q.
• Micron fiscal Q4 announcement, August26.
• Prior weekly calls, preserved for historical grading; new targets point to the September 6 campaign map.

Predecessor reconciliation

All eleven September 4 live targets and eleven older same-name live instructions are CLOSED as superseded (22 total). One older MU watch was already explicitly closed on August 28; its missing machine resolution is restored to August 28, not regraded September 6. The July8 IREN stop/invalidated result retains its July17 resolution. No unrelated target is resolved by this campaign.

The July NBIS trigger fired August12 but skipped its intended entry; its annotation is preserved without inferring a fill. The older NVDA technical leg was MET August8, without proof of its original full capex conjunction. The July MU peer-print conjunction, SNDK additional-print durability, WDC allocation/margin/capacity language, and CEG July capex-language questions remain unresolved evidence conditions. New price levels do not erase those questions.

Every historical source retains its original machine stance for the publication-date scoreboard. Each receives only a visible, named supersession notice. Every new entry target links its predecessors. The twelve reviewed older-target capsules are archived beside the diagnostic; cited source articles were not supplied to the clean analyst.