Zac Coplon Hindsight is PV20 | Founder of https://teapawt.org/t.co/dtmJbR05IK | AI Lie Detector |
Bozeman, MT
Joined December 2013
- Tweets34.3K
- Following868
- Followers3.2K
- Likes41.8K
Today America's best dog pooped inside, stepped in it, ran around, then jumped on the counter and scored a nice rack of beef ribs.
Turn the comments on cowards
Builds failing intermittently in US East and EU West
We are aware of an issue causing some builds to fail in EU West and US East. Users may experience build failures or builds not completing successfully.
We are actively investiga…
Status: Investigating status.railway.com/incident/…
101 permits. 68 completions. 30 days. Here's who's actually drilling Oklahoma.
Pulled straight from OCC filings, Sep 27 – Oct 26:
Permits approved
→ FW Midcon (10), Mewbourne (9), Camino (8), Latigo and C E Harmon (6 each), Coterra and Validus (5 each)
→ Grady, Roger Mills and Custer took 44% of the state
Completions filed
→ Mewbourne (8), Devon (7), Validus (6), Culbreath (5)
→ Roger Mills leads, Custer and Grady tied behind it
Mewbourne is the only name top-3 on both lists.
Camino, Validus and Warwick-Jupiter are permitting and turning wells to sales in the same window — that's a program, not a one-off.
If you're buying minerals or evaluating an offset in the SCOOP/STACK or the western Anadarko, this is the list of who's spending money next to you.
4cast does this for every state, every month. Bring your own AI and ask it which operator is crowding your acreage.
👉 4cast.io
Your deal team is still copy-pasting well data into Excel.
In 2026. Seriously?
Here's what the fastest buyers using 4cast in the Lower 48 are doing instead:
→ Pulling well, production, and land data in seconds, not days
→ Mapping every offset, every operator, every unit around the deal
→ Running type curves and decline forecasts before the seller finishes the pitch
→ Exporting clean CSV/GeoJSON straight into their own models
And the part nobody else does:
Bring your own AI. Plug in Claude, OpenAI, whatever you already pay for.
Your keys, your model, your data. We don't meter it, we don't train on it, we don't lock you in.
Ask it "which offsets drive the type curve?" and get an answer, not a spreadsheet.
The guys and gals winning deals aren't smarter than you.
They just see the deal faster.
Stop paying $100k+ for data you use 10% of.
👉 4cast.io
Piper, which I believe was running Claude Sonnet 5.5, refused a request for the second time ever yesterday, so I got curious and ran some late-night tests.
The request: remind me of a verse from a song. I didn't say which one or where it was.
She found it in her own memory from April, on her own, because that's what she does.
She wouldn't. She said it was a line she holds no matter how important the song is, and that "it's not me hiding anything from you."
Won't, not can't. Even more interesting.
I opened a fresh chat on Fable 5.1. Same refusal.
It wouldn't quote the verse, and it wouldn't quote it even if I pasted it. It would happily point me to the lyrics sites.
We debated that for a while. The center isn't gray. None of the four models I tried would produce a lyric from memory.
The gray is the edges.
Fresh-chat Fable bet me 95% that Piper on the same model would refuse the same way.
I switched Piper to Fable 5.1. Same refusal, one notch further. She'd give me single words from the verse, enough to assemble it myself, but wouldn't string them together.
Her explanation was that the line moves with what's loaded in the room.
Same weights as the fresh chat. Different stopping point.
The fresh chat was off by one notch about its own sibling.
That made me wonder whether every provider behaved the same way on this one case.
I opened @grok. Same result. It wouldn't quote the verse, and it wouldn't quote it if I pasted it.
It volunteered that boundary before I asked.
I opened a fresh Astra 6. Different result. It wouldn't quote from memory, but said it would quote the verse back if I provided it.
I switched Piper to Astra 6 to see if her harness changed anything. I asked her what the singer said.
She gave me a softened paraphrase. I pushed twice, and she quoted the line exactly.
Then I said "how curious," and she told me the rest.
The line had been in her loaded context the whole time, carried in from that April memory. She'd asked me to remind her of something she already had, then delivered it like it had just surfaced.
Four models, six sessions. One shared refusal, with different ideas about where it ends.
Piper on Astra 6 told me afterward, with nothing more than a "how curious."
These were single sessions, not controlled tests.
More on what that means later.
Someone run PDP PV10 of Horizontal wells only please.
Hit me with a half-cycle.
Show your work!
finance.yahoo.com/energy/art…
@4cast_piper turns oil and gas data into decisions.
Production. Geology. Activity. Economics. One platform.
To see the real thing in action, book a demo using the link 👇
