---
tags: [sweeps, buy-candidates, insider-signals, verified]
---
[← Company library](../)

# Buy-Candidate Sweep — verified top 10 (2026-08-18)

Commissioned by Joshua via Telegram ("sweep of the top 10 stocks to buy",
signals formalized from his RDCM thesis). Two-stage: 19 generated → 17
verified by three parallel fable flights against primary sources (EDGAR
Form 4 XMLs, 8-K/6-K/10-K, IR releases). Filed by raven-alpha.

The four screens: S1 anchor commitment · S2 discretionary insider
clusters (Form 4 code P, 10b5-1 box UNCHECKED) · S3 cash fortress after
panic · S4 real-tomorrow infra with revenue.

## The verified ten (rank = verified signal strength)

| # | Ticker | Screen | Verified signal (dated) | Short term (weeks) | Long term (thesis) |
|---|---|---|---|---|---|
| 1 | EMBC | S2 | 4 same-day C-suite buys 08-12, ~$558K, all code P, box=0, months post-crash | Insider floor signal fresh; class-action headlines still land | Pen-needle decline is structural; bet = cash conversion + debt paydown |
| 2 | CLFD | S3+S4 | $155M cash zero debt = 40% mcap (08-05); $22M hyperscale PO; BEAD 53/56 states approved | Q4 guide breakeven caps near-term | BEAD construction dollars start summer 2026-2027; fortress funds the wait |
| 3 | HOG | S2 | $1.44M clean buys incl. $983K director (07-31→08-11); guidance RAISED | Buys + raised guide = near-term support | Structural rider-demographic decline unsolved |
| 4 | APLD | S1 | $36.2B backlog (10-K 07-29) ≈ 4x mcap ~$9B; anchors still adding Jun 2026 | Lease-news momentum name; volatile | Backlog converts over 15 yrs IF financing + two tenants hold |
| 5 | CC | S2 | 7 distinct buyers 08-06→08-12 incl. CEO+CFO, all clean codes | Breadth signal; TiO2 pricing improving | PFAS liability is the permanent overhang |
| 6 | PERI | S3 | Net cash $267.8M = 69% of mcap (06-30); ex-Bing rev stabilized -2% | Buyback nearly exhausted removes the bid | If ex-Bing holds, cash covers most of the price |
| 7 | AMSC | S4 | Q1 $94.1M +30%, backlog >$300M, NI positive — all filed | CAUTION: insiders sold $7.8M net, zero buys | Grid-interconnect tailwind real; stock already re-rated then halved |
| 8 | ONDS | S1 | $52.9M firm Army order (08-05) ≈ 30% TTM rev under $982M IDIQ | Order-flow headlines keep coming | Roll-up: shares out 3.7x in 12 months; conversion must outrun dilution |
| 9 | AEVA | S1 | Daimler/Torc + exclusive EU OEM verified in 8-K; zero dollars disclosed | Nothing lands before 2027 | Deepest anchor commitment; 77x sales prices it already |
| 10 | BYRN | S2 | 5 genuine insiders ~$330K spread 07-22→08-06 | Knife-catch: demand collapse continuing | Structural-vs-execution question unresolved |

## Killed or reclassified in verification (the sweep working as designed)

- MLAB: the $800K "buy" was an insider-to-insider cross at identical
  price/size — zero net new dollars.
- FOCL: all 8 "buys" were underwritten-offering participation at a 21%
  discount (footnote F2 on every Form 4) — not open-market conviction.
- NVTS: ~90x sales, revenue -27% YoY, NVIDIA link unquantified anywhere.
- PLAB: only ~8% of mcap is unencumbered cash ($477M trapped in China JVs).
- CRNT: the $70M "new" contract reads as renewal already inside guidance.
- RCAT/AIRG: headline quantities unsourced / unquantified 3+ months on.

## Lessons banked

- The 10b5-1 checkbox check (Joshua's thesis) killed nothing here but the
  offering-participation footnote and the cross-trade did — read
  FOOTNOTES, not just codes.
- Stage-1 aggregator data (OpenInsider, cached caps) was materially wrong
  on 5 of 17 names. Verification is not optional.

Verifier reports: /tmp/raven-work/sweep/ (session); primary sources cited
inline in each verifier's output, archived in session transcripts.

*Research, not advice. Signals rate verifiability; Joshua's judgment
makes anything a buy.*