Crude Cut № 001 · GreenLake Energy · Delaware basin, TX
Two miles, then one, then three. GreenLake Energy — formed by Matt Gallagher after the $7.6 billion sale of Parsley Energy, backed by NGP — bought scrap acreage in the Delaware basin and has drilled it in three acts: standard two-milers, then a short-lateral test and program, now horseshoes. Each act is measured the same way: median cumulative oil per well, GreenLake vs comparable wells within ten miles.
Act one · 2022–23 · two-mile wells
37 wells spudded in 2022–23 across Pecos, Reeves, and Loving counties, nearly all full two-milers. Against the 427 two-milers within ten miles of their pads, they run even for the first year and finish 4% ahead at 24 months. The acreage also came with short Wolfcamp XY wells drilled by prior operators in 2012–2019; in early 2023 GreenLake added three one-mile tests of that bench to an eight-well pad — the next panel.
Median cumulative oil per well · first 24 months online
RESULT: EVEN WITH OFFSETS. 157,235 vs 157,000 bbl at 12 months; 235,090 vs 226,170 at 24 (n=32 vs 427) — +4%.
The test · FEB–MAR 2023 · JAYHAWK block, Loving Co.
The Loving leases came with prior operators' wells on them, including one-mile Wolfcamp XY horizontals drilled 2012–19. The best of those made 38,735–39,545 bbl per 1,000 ft in year one, against 19,333 for the area's offset short wells; the ground's one full-length Wolfcamp B Upper well made 3,813. In FEB–MAR 2023 GreenLake drilled eight wells on the JAYHAWK block: five 10,500-ft Wolfcamp B Upper laterals and three 5,100–5,300-ft Wolfcamp XY laterals on tracts too small for full-length wells, two of them filed as allocation wells. The shorts came online OCT 2023 – FEB 2024. The first well of the 2024 short program spudded APR 2024.
The JAYHAWK block · inherited wells, the 2023 pad, and the wells that followed
inherited — drilled 2012–19 the 2023 test — three XY shorts same-pad 2-milers GreenLake 2024–26 other operators (2,288) · equal-mile · hover to list wellbores
The test, month by month · cumulative oil per well · first 24 months online
RESULT: THE WORST TEST BEAT THE PAD'S OWN 2-MILERS, PER FOOT. Year one, per 1,000 ft: the three tests made 48,200 / 41,445 / 26,907 bbl vs 24,413 for the median same-pad 2-miler (the five ranged 7,392–30,123). Reference: the 397 offset short wells online 2015–22 within 10 mi ran a median year one of 100,190 bbl (19,333 per 1,000 ft), 90th percentile 177,180. JAYHAWK 65 4002H made 247,360 bbl from 5,132 ft — above that p90, and above GreenLake's own 2-mile median (157,235).
The tell · 2024 · lateral mix · Reeves · Loving · Pecos · Ward
One-mile wells weren't a secret — fifteen other operators spudded at least one in 2024. They just weren't a program: everyone else in the four counties combined drilled 64 short laterals against 698 two-milers, a six-percent sideline. GreenLake put 15 of its 22 spuds into one-mile wells — more one-milers than any other operator in the four counties, majors included.
Share of 2024 spuds that were one-mile wells (3,500–7,000 ft) · spud year 2024 · four-county TX Delaware
one-mile share — GreenLake one-mile share — everyone else the rest of that group's 2024 spuds · hover a ring for the full mix
RESULT: THE PIVOT HAD NO COMPANY. Everyone else's 64 one-milers came from 15 operators drilling them as a sideline — Mewbourne 11 of 32 spuds, Petro-Hunt 10 of 37, EOG 8 of 72, Chevron 7 of 36, Occidental 5 of 204. The only comparable mix in the basin was Circle-S Energy, a ten-well program (6 of 10). GreenLake's balance: four two-milers, all spud by FEB 5, and the first three horseshoes, spud OCT–DEC.
Act two · 2024 · one-mile wells
Seven months after first test flowback, the 2024 program moved to one-mile wells: fifteen, aimed at Wolfcamp XY and the A bench, drilled in whole-tract batches — six on SODA POPINSKI, five on MEDICINE BOW (where the inherited 2017 DUC was also completed), four on MACHO MAN. Year one, on the same ground and vintage, their shorts made 42% more per foot than offset short wells.
Median cumulative oil per well · first 12 months online
RESULT: 70% OF THE OIL FROM HALF THE HOLE. Year-one median 133,520 bbl vs 190,590 for offset 2-milers — from half the footage. Per foot: 29,265 bbl/1,000 ft vs 20,550 for offset short wells (+42%) and ~19,250 for the 2-milers (n=13 vs 60 vs 347).
Act three · 2025–26 · horseshoes, cubes, the whole stack
45 wells spudded since January 2025, across four programs at once. Eighteen Bone Spring wells, drilled back across the early ground (AD HAWK on the first JAYHAWK unit; a 17,749-ft lateral — their longest filed — on the original RICK ROSS lease). GODZILLA: eight wells stacked across four benches on one Reeves unit. Two Wolfcamp C tests. And horseshoes — two one-mile legs turned 180° into one wellbore: six online, ahead of the straight 2-milers around them, with two six-well 3-mile rows (EL HEFE, EL CRUDERINO) waiting on completion, funded by a further $200MM from NGP.
The 2025–26 program on the ground · every GreenLake well since 2022
spudded since JAN 2025 horseshoe (U) well 2022–24 wells other operators within 2 mi (1,911) · laterals drawn to scale · hover a pad to list its wellbores
Median cumulative oil per well · first 8 months online
RESULT: HORSESHOES +35% AT 8 MONTHS. Median 166,615 bbl vs 123,150 for offset straight 2-milers of the same vintage (n=6 vs 165). The 3-mile rows and the GODZILLA cube are not yet scoreable at the JUL 2026 cut.
The tell, again · OCT 2024 – JUL 2026 · horseshoe share of spuds
The one-mile pivot wasn't a one-off — it was a method, and it's running again. Since OCT 2024, when their first horseshoe spudded, GreenLake has put 9 of its 51 spuds into U-turns — one well in six. The rest of the four counties: one in twenty-five. The difference this time is the company: nobody copied the shorts in 2024, but the horseshoe queue behind GreenLake is now 14 operators deep — and GreenLake's six online are the ones with a year of receipts.
Share of spuds that were horseshoes · OCT 2024 – JUL 2026 · four-county TX Delaware
horseshoe share — GreenLake horseshoe share — everyone else the rest of that group's spuds · hover a ring for the full mix
RESULT: THEY'RE DOING IT AGAIN. Horseshoes were 17.6% of GreenLake's spuds in the window (9 of 51) vs 4.1% for everyone else (72 of 1,770). A horseshoe here = a U-suffix well name, or a filed 9,000–12,500 ft lateral whose endpoints sit 400–3,000 ft apart. Fine print: GreenLake wasn't first online — Permian Resources and APA each had one flowing by JAN 2025 — but at the JUL 2026 cut GreenLake's six producing horseshoes were the basin's largest cohort with receipts; the 55 spudded behind them were mostly still waiting on first production.
Rebuild this cut
Every number on this page traces to a query that runs verbatim in a Crude Code-connected Claude session, against the same public state filings. Pull the recipe, re-run it on current data, change the cohorts, or point it at your own counties.
crudecode.devThe measure: median cumulative oil per well by months online, balanced panels — every well in a cohort has the full window reported. The analogs are spatial offsets: wells of other operators within 10 miles of a GreenLake well of that act (surface locations, ST_DWithin), same footage class, same vintage. Texas RRC filings via the Crude Code DB, JUL 2026 cut.
Act one: GreenLake = 2022–23 spuds, 8,500–11,500 ft lateral (n=32 with 24 months). Offsets = same band, first production JUL 2022 – FEB 2024, within 10 mi (n=427). Act two: GreenLake = 2024 spuds, 3,500–7,000 ft (n=13 with 12 months; 15 drilled). Offsets = same band, first production JAN 2024 – JUL 2025, within 10 mi (n=60). The dashed reference line = offset 2-milers, same definition at 8,500–11,500 ft (n=347); "70% of a 2-miler" = 133,520 ÷ 190,590. Per-foot = year-one cum ÷ filed lateral × 1,000 (29,265 vs 20,550, and ~19,250 for the 2-miler reference). Act three: GreenLake = horseshoe (U) wells online (n=6, first production MAY–NOV 2025). Offsets = straight 8,500–11,500 ft Reeves wells, first production in 2025, within 10 mi (n=165). A horseshoe files both legs as one lateral (~10,200–11,200 ft), so straight 2-milers are the footage-matched comparator. Act-three scope = 45 wells spudded JAN 2025 – JUL 2026; "18 Bone Spring" is the RRC-filed formation count in that class. The 2023 test: three one-mile Wolfcamp XY wells (5,132–5,322 ft) drilled FEB–MAR 2023 on the JAYHAWK block in the same batch as five 10,475–10,603-ft Wolfcamp B Upper wells; two shorts filed as allocation wells; first online OCT 2023. Year-one per 1,000 ft = first 12 months' cum ÷ filed lateral × 1,000: tests 48,200 / 41,445 / 26,907; same-pad longs median 24,413 (range 7,392–30,123). Through JUL 2026 the tests stand at 340,200 / 335,890 / 228,430 bbl in 33 / 29 / 29 months; the first online made 247,360 bbl by month 12 and 310,700 by month 24 — clear of the act-one 2-mile medians (157,235 / 235,090). The 2024 short program's first spud followed APR 2024, within seven months of first flowback. What they knew going in: the inherited 2012–19 wells on these leases ran Wolfcamp XY up to 38,735–45,690 bbl/1,000 ft year one (MOORE HOOPER A 3H, HAMILTON SE 12H, MORLEY NW 11H; the 2019 HAMILTON N 11S short: 39,545) against one B Upper long at 3,813; the neighborhood's short-well record — 397 offset shorts (3,500–7,000 ft, other operators, first production 2015–22, within 10 mi of the test wells, full first year reported) — ran a median 100,190 bbl year one (19,333 per 1,000 ft), 90th percentile 177,180. The test map: every producing GreenLake wellbore in the Loving block by role, over other operators' laterals in the window (n=2,288), equal-mile, no basemap. The map: filed lateral wellbore geometries for every GreenLake horizontal spudded since 2022, drawn to scale (green = GreenLake: solid dark = spudded 2025–26, deepest = horseshoes, dashed = 2022–24); background = every other operator's lateral within ~2 miles of a GreenLake well (n=1,911), drawn heel-to-toe. Equal-mile projection; no basemap. The lateral-mix census ("the tell"): spud counts, spud year 2024, Reeves–Loving–Pecos–Ward — county scope, not 10-mile offsets. One-mile = 3,500–7,000 ft filed lateral; two-mile = 8,500–11,500 ft; horseshoes by U-suffix well name; "everyone else" = all other operators' 1,040 spuds (37 operators, 15 of them with at least one one-miler). The horseshoe-share census ("the tell, again"): same county scope, spuds OCT 2024 (GreenLake's first horseshoe spud) through JUL 2026. Horseshoe = a U-suffix well name, or a filed 9,000–12,500 ft lateral whose endpoints sit 400–3,000 ft apart — two one-mile legs turned 180°. 17.6% = 9 of 51; 4.1% = 72 of 1,770; "4× over-indexed" = 17.6 ÷ 4.1, or equivalently 11% of the window's 81 horseshoes on 2.8% of its 1,821 spuds.
What we'd push back on ourselves: six horseshoes and thirteen one-milers are small samples — medians, not miracles. The 2023 test conflates bench and length — the shorts are Wolfcamp XY, the pad's longs are B Upper — so it answers the commercial question (what does a scrap tract earn?) rather than a clean length-normalization test. Against county-wide cohorts instead of 10-mile offsets, act one reads 90% of the field — that gap is county mix (their program is Reeves/Pecos-heavy), which is why we use offsets. The short-bench idea wasn't conjured from nothing: prior operators drilled short Wolfcamp XY wells on these leases in 2012–2019, so GreenLake's edge is completion-era execution on a concept the acreage came with. "Pay," in act two's headline, is an inference — state filings carry no cost data; what the filings show is oil: 70% of a 2-miler's cum from half the footage, +42% per foot. Horseshoe identification is convention-dependent: U-suffix naming isn't universal, and some filed geometries are degenerate (15,000-ft laterals whose filed endpoints coincide — including two of GreenLake's own GODZILLA U-wells), so the census cross-checks name and geometry and still undercounts wells with missing or bad geometry. Corporate facts (NGP commitments, the Parsley sale) are from public announcements, not state data.