Let's Ask Claude

The short version

In Dispatch No. 16 I showed the very obvious typo (or worse?) that resulted in a deeply flawed allowance.

Let's see if we can uncover any other errors.

This is a long, boring one… But it uncovers some very interesting structural flaws in the patent system. These are things that are currently being exploited, and you won't realize what's happening until it's too late (it's already too late). It's not even what happened to me. This post is more of a cautionary tale, wrapped up in a bunch of armchair analysis (you all know I am not qualified to comment on any of this, so feel free to set me straight).

One important caveat before we dig in, and it's the same point I made back in Dispatch No. 2: an examiner's mistakes — even glaring, self-contradicting ones — are not themselves grounds to cancel a patent. A PGR stands or falls on §§ 101/102/103/112, not on “the examiner was sloppy.” So read most of what's below as context — how a bad claim slipped through, and how much benefit of the doubt anyone's earned — rather than the legal case itself. That said, one of these findings doubles as real evidence on the merits, and I'll flag it when we get there.

The prompt

Review the USPTO prosecution-history PDFs (in the current folder. see the First Office Action, Applicant response, Final Rejection, Applicant RCE response with amended claims, and Notice of Allowance). Analyze the examiner's office actions for technical errors or unresolved contradictions under the MPEP or statute without assuming any knowledge of the prior art's actual contents. Pay particular attention to internal contradictions in the record—especially whether a limitation the examiner relied on to allow a claim was earlier found anticipated by the same reference, and whether the stated reasons for allowance actually match the text of the claims they allow. Also carefully compare the applicant's remarks against the true text of the amended claims to flag any typos or misleading statements in the amendment and remarks.
Flag argument remarks that assert factual statements (not opinions) characterising the prior art.
Limit output to 2 sentences per finding, citing relevant MPEP or statutes.
Doublecheck your analysis.

These files are in downloads.

What Claude found

Remember this Notice of Allowance was July 2025. Claude Opus 4.1 was released August 5, 2025. It's hard to find old models still active. I objectively cheated here by using Opus 4.8.

Output:

Interleaved with my commentary

A. Dispositive contradiction: a limitation relied on for allowance was twice held anticipated by the same reference

[Claude Output]
A-1 — The sole stated reason for allowance is a limitation the examiner had already found anticipated by Phillips.

The Notice of Allowance allows "Claim 1 and all of it[s] dependencies" because "Puwanto, along with all other references, fail to teach continuing sampling the rotational values through a second direction change … and ceasing sampling the rotational values at a third direction change" (NOA p.5), yet both the non-final action (p.5) and the Final Rejection (p.6) expressly held that exact limitation — former claim 7's "continuing through a second change of direction … and ceasing at a third change of direction" — "Phillips further teaches," citing Phillips ¶¶[0029],[0110]. A reference cannot simultaneously anticipate a limitation and "fail to teach" it, so the reasons for allowance are irreconcilable with the rejections of record (35 U.S.C. 102(a)(1); MPEP 2131; MPEP 1302.14).

The examiner read my application (but probably only once). Here's the first OA:

[First OA, page 5-6]
Regarding Claim 7, Phillips further teaches the sensor system of claim 5, wherein the processor subsystem is configured to continue sampling the rotational velocity along substantially an entire stroke of the at least one rod, the sampling beginning at a first change of direction of the at least one rod, continuing through a second change of direction of the at least one rod, and ceasing at a third change of direction of the at least one rod [0029; 0110].

I'd have cited different paragraphs (there's a lot more content in my application), but still. Right there. The examiner stated that I teach that exact thing that claim 7 states (which is correct, I do teach that).

Then again in the Final OA/Rejection

[Final OA, page 6]
Regarding Claim 7, Phillips further teaches the sensor system of claim 5, wherein the processor subsystem is configured to continue sampling the rotational velocity along substantially an entire stroke of the at least one rod, the sampling beginning at a first change of direction of the at least one rod, continuing through a second change of direction of the at least one rod, and ceasing at a third change of direction of the at least one rod [0029; 0110].

This was cut and pasted from the First OA, but it was a second assertion on the very same fact. You have to count this as twice. It was still a § 102 when maybe it should have been a § 103… but cutting and pasting is way easier. Also, there was an examiner error in this same section by stating rejected claims 13-15, but then citing 19-20 (see Finding B-3 below). The examiner was clearly confused about claim 13 all the way back to the Final.

[Claude Output]
A-2 — The relocation that produced the contradiction is documented in the amendment markup.

The RCE listing strikes that entire "first/second/third change of direction" clause from claim 7 and adds a reworded version into independent claim 1 (RCE claims: claim 7 struck-through; claim 1 newly-underlined) — i.e., the precise text earlier mapped to Phillips ¶¶[0029],[0110] became the basis for allowance, unchanged in substance ("rod"→"component," "velocity"→"values"). The contradiction is sharpened because the same NOA cites Phillips "[0026; 0029; 0043; 0096; 0134; 0138; Fig. 6]" for claim 1's rotational sampling while declaring no reference teaches the second/third-direction-change sampling that ¶[0029] had been cited to anticipate (MPEP 1302.14; MPEP 2131).

I'd disagree with Claude here on ("rod"→"component," "velocity"→"values"): those are significant broadenings of the terms. But it's right on the text of claim 7 being twice previously found unpatentable by the examiner (under §102 but sadly not §103). It'd be different if the examiner indicated that was allowable, but that wasn't the case, twice prior. And using the same paragraphs twice to say I disclose it, then those same paragraphs again to say I don't, with no unwinding explanation?

[0110] … The magnitude of the rotational acceleration, velocity, and finally displacement can be measured through the use of an accelerometer and a gyroscope over a single stroke. These measurements, in turn, can be compared over many strokes to establish a torque buildup and release period and magnitude.

I'm talking about a very specific type of rotation here. The crux is whether I specifically use the magnetometer to bound the stroke. The short answer is there are many ways to bound the stroke. The answer in this context is paragraph [0082] (see below) where I talk about the different ways to bound the stroke, which includes the magnetometer as one option. This doesn't have to be a core element, it just has to be in there, and it very clearly is. Just in a paragraph that the examiner apparently didn't get to.

• First OA
• Final OA
• Amended Claims

My [0029] was cited, but I'd have used my [0082]. Here's what those paragraphs state (note, [0082] is a much stronger text in this context):

[My disclosure]
[0081] The sensing device 201 comprises a sensor. For example, the sensor may be a magnetometer that senses magnetic field. Magnetometers operate by reading the strength of a magnetic field in a given direction. The magnetometer is effectively a compass that identifies a radial orientation of the polished rod relative to a fixed external location such as magnetic north. In the present context, the magnetometer is used to detect a change in radial orientation of the polished rod. After calibration, a three-axis magnetometer can determine direction, or compass heading, of polished rod within the magnetic field.

[0082] The sensor may further comprise an accelerometer, a gyroscope or a barometer. An accelerometer can be used to detect the accelerations imparted by the directional reversals caused by stroking action. A 3-axis gyroscope is similar to a 3-axis accelerometer, except the frame of reference is a static inertial frame. A barometer can measure changing air pressure indicating the rise and fall of the sensing device. These sensors aid in identifying vertical, cyclic motion of the polished rod 101, and thus indicate that the pumping unit is stroking and rotation is accordingly expected. The magnetometer experiences a range of field distortions as the sensor moves closer and further from steel objects in the pumping system. Short term variation in the magnetometer readings can therefore also be indicative of stroking action.

That's it! The examiner said I teach this, then turned around and believed the applicant's claim that I never said this exact thing in this exact way — even though that same examiner had stated it was taught, twice. You're probably saying "oh, but you're calling out the accelerometer, not the magnetometer", but read the 2 paragraphs together. Read the last 2 sentences that I emphasized because that's super important!

And here's the flag I promised — the part that isn't just “the examiner got confused.” In his own words, twice, the examiner wrote that Phillips further teaches claim 7's first/second/third change-of-direction sampling, citing specific paragraphs. That's not my characterization; it's the examiner's official finding, on the record, that the prior art already teaches the very limitation later blessed as the sole reason for allowance. An examiner's own anticipation finding doesn't prove a § 102 case — the Board decides anticipation fresh, from Phillips's actual text, and that text is the evidence that does the work. But it's a powerful piece of the record: the Office, in its own words, twice mapped this very limitation onto Phillips. The contradiction embarrasses the allowance; the underlying admission helps arm the petition — the prior art itself does the rest.

B. Internal contradiction within the Final Rejection (vibration limitation)

[Claude Output]
B-1 — Claims 13–15 are listed as anticipated by Phillips and, in the same action, rejected as patentable over Phillips for a feature Phillips admittedly lacks.

The Final's §102 heading states "Claims 1-7 and 13-15 are rejected … as being anticipated by Phillips" (p.4), but the body rejects claims 13–15 under §103 over Phillips in view of Puwanto because "Phillips fails to teach a vibration sensor subsystem for monitoring vibration … in three axes" (pp.13–15). A claim cannot be both anticipated by a reference and separately rejected as obvious for an element that reference fails to teach (35 U.S.C. 102/103; MPEP 2131, 2143.03, 706.02).

Typo? No! (well, yes, but it's worse than a typo) On page 4 of the Final OA it rejects 13-15 under a §102, but in that section it talks about claims 19-20. So which is it? Later on page 12, it rejects claims 12-15 (which includes 13-15) under a §103.

For clarity, let's look at the First OA since they were largely cut and pasted. In this one 13-15 is squarely in a §102 (both header and body). In the Final OA, 13-15 (plus 12) was moved to a §103.

So now ask, what was the typo in the Final OA? The header or the body of the §102?

I can't let the vibration sensor thing go. The examiner was just wrong here. I also didn't give Claude the source document, but this is especially interesting because even assuming it to be true (which it isn't), Claude uncovered the technical contradiction. I hate harping on this, but I couldn't have more clearly provided this 3-axis vibration sensor in the disclosure. It's a poorly sketched figure, but straight to the point. I use this same figure to reference the accelerometer (by definition a vibration sensor), a gyro, which I repeatedly describe as sensing "rotational oscillations" (oscillations are vibrations). And you could easily argue the stroking motion of the system is a low-frequency vibration (that sounds like a stretch, but it really isn't). That means I'm using the magnetometer as a vibration sensor, also in three axes. And the barometer. Have you ever heard of a barometer being used as a vibration sensor? Yeah, that's exactly how I'm using it too (but that's only in one dimension).

[Claude Output]
B-2 — The §102 anticipation of independent claim 1 never addresses claim 1's vibration element.

As it stood at the Final, claim 1 recited "a vibration sensor subsystem for monitoring vibration … in three axes" (shown unmarked — i.e., pre-existing — in the later RCE listing per 37 CFR 1.121), yet the Final's claim-1 §102 mapping accounts only for the magnetometer, gyroscope, and processor and is silent on the vibration element the same action says Phillips lacks. An anticipation rejection must account for every claimed limitation; omitting one while elsewhere admitting the reference does not teach it is an internal defect (MPEP 2131; 35 U.S.C. 102(a)(1)).

I'm still annoyed about the whole vibration sensor thing, but this is interesting. Yet another indication of the examiner's lack of attention. Maybe "variations in the magnetic field" counts… but wait, isn't that sort of like a vibration sensor (yes and that's how I use the magnetometer in one aspect), but more importantly the element of the vibration sensor in claim 1 was addressed in neither the First OA nor the Final OA. That was addressed in claim 12, but not claim 1! That's except for acknowledging in claim 1 that the magnetometer can detect variations (if that's not the definition of vibration sensor, what is?)… and then the examiner goes on to contradict that in claim 12 by saying I don't teach a vibration sensor 🤯

[Claude Output]
B-3 — The §102 heading misidentifies the rejected claims (notice defect).

The Final's §102 heading lists "1-7 and 13-15," but the body actually rejects claims 19–20 under §102 (p.6–7, "Regarding Claim 19, Phillips teaches a method…") and moves 13–15 to §103 — the heading appears copied verbatim from the non-final action without updating. Applicant's own RCE remarks had to re-state the ground correctly as "Claims 1 through 7, 19, and 20," confirming the heading failed to give clear notice (35 U.S.C. 132(a); MPEP 707.07).

🍌🍌🍌Bananas!!!🍌🍌🍌 The examiner is making stupid mistakes. I didn't even catch this one. Thanks Claude. Yep, it's right there. The header doesn't match the text and vice versa. How is claim 13 so frickin' slippery? It's like the Teflon claim.

This really makes me wonder if the applicant caught this and compounded it by that "innocent typo." I don't think it was innocent, but proving that is hard. See Dispatch No. 16 — Claim 13 — Again

C. Reasons for allowance vs. the actual allowed claim text

[Claude Output]
C-1 — The reasons for allowance do not match allowed claim 13.

The NOA allows claims 13–15 only by asserting claim 13 has "analogous, though not necessarily coextensive, features … for similar rationale" as claim 1 (p.5), but amended claim 13 contains neither the "second/third direction change" sampling limitation that is the stated reason for allowance nor (after the RCE struck them) the vibration/vibrational-baseline limitations — its only distinguishing feature is comparing "rotational values with previously sensed rotational values … during a previous stroke." Allowing claim 13 by reference to a limitation it does not recite leaves it without a supported, claim-matched basis for allowance (MPEP 1302.14).

This one is self-evident. See Dispatch No. 16 — Claim 13 — Again. It is fascinating that Claude processed these sequentially (C-1 through D-2).

D. Applicant's remarks vs. the true text of the amended claims (typos / misleading statements)

[Claude Output]
D-1 — Wrong claim number in the §102 argument.

Under the heading "Claims 1 through 7," the RCE remarks argue "Phillips does not describe the sensor system of amended claim 13 including a processor subsystem to: detect axial movement … sensing a direction change …" (RCE Remarks p.1) — but those are claim 1's limitations, so "claim 13" is a mislabel for claim 1 (cf. amended claim 1; 37 CFR 1.111(b)).

See Dispatch No. 16 — Claim 13 — Again. It's amazing how complex this was and how Claude was able to untangle it with just the raw documents.

[Claude Output]
D-2 — Mislabeled "independent" claim.

The RCE remarks assert Phillips/Hurst do not "describe each and every element of independent claim 18, as currently amended" (p.3), but claim 18 is dependent on claim 16; the next sentence (arguing "amended claim 16") shows "claim 18" should read claim 16 (37 CFR 1.111(b)).

The applicant RCE remarks

Ok, I'll call this a legitimate typo. Maybe. At least this is within the chain of dependency, unlike that whole Claim 13 thing.

[Claude Output]
D-3 — Mischaracterization of the Final Rejection.

The RCE remarks state "The Final Office Action points to various generic statements in Phillips" regarding strokes/sensing (p.2), whereas the Final made a specific anticipation finding mapping the first/second/third-change-of-direction sampling to Phillips ¶¶[0029],[0110] (p.6) — the characterization understates the record the applicant was traversing (MPEP 2145).

The RCE

You can see the applicant getting bolder and more assertive with the mischaracterizations. A line was crossed somewhere here, but I'm not qualified to identify where.

E. Argument remarks asserting factual statements about the prior art (flagged)

[Claude Output]
E-1 — Categorical factual assertions contradicted by the examiner's own findings.

The RCE remarks assert as fact: "there is no disclosure in Phillips that the sensing then continues during a second change in direction" and "no disclosure in Phillips of then ceasing the sensing once a third change in direction is detected" (p.2) — flat factual characterizations of the reference that are the exact opposite of the examiner's record findings (Phillips ¶¶[0029],[0110]); the examiner adopted them in the NOA without reconciling his contrary prior position (MPEP 2145; 1302.14).

This is what I'm talking about in Dispatch No. 12 where I, as the inventor and with the most to say and lose, am silenced on the sidelines. For F&*#$ sake, just read the prior art submission. That's all you need to find the citations. Of course the examiner is going to believe the applicant. It's a one-sided conversation.

[Claude Output]
E-2 — Other flagged factual characterizations of Phillips.

Non-opinion factual assertions about the reference include the non-final remarks' "Phillips … '[t]he magnetometer is effectively a compass that identifies a radial orientation …' ¶[0022]" and "As recognized in the Office Action, Phillips fails to teach a vibration sensor subsystem. Office Action, page 11" (M171 pp.2–3, the latter accurate to the non-final action p.11), plus the RCE remark "while Phillips discussed some comparison of values, Phillips does not disclose using the value detected during a previous stroke" (RCE Remarks p.5) — all assertions of fact about the reference's contents rather than argument (MPEP 2145 / 716.01(c)).

Claude likes to shorthand filenames — M171 pp.2–3 is the applicant response to the first OA and it's really on page 3.

As recognized in the Office Action, Phillips fails to teach a vibration sensor subsystem. Office Action, page 11

Claude doesn't even know this is incorrect, because I didn't provide the source document (i.e. my application). It's crazy how it caught this even if it was for an unrelated reason.

In my opinion, this is misleading advocacy of the sneakiest kind. Notice how carefully it's worded: the applicant doesn't assert the false thing outright — they quote the examiner saying it (“As recognized in the Office Action…”), so the falsehood rides in on the examiner's own authority while the applicant keeps clean hands and we all get it psychologically reinforced, wink, wink. That careful construction is exactly why I'd call it misleading rather than a flat-out lie — and again, all my opinion. I'm open to hearing yours.

F. Minor / drafting items the examiner did not address

[Claude Output]
F-1 — Double-introduced element (possible antecedent-basis / §112 issue).

Independent claim 1 recites "a vibration sensor subsystem … in three axes," and dependent claim 12 then says "further comprising a vibration sensor subsystem … in three axes over a vibrational baseline" (RCE claims, claims 1 & 12), redundantly introducing a second such subsystem rather than referencing the one in claim 1 (35 U.S.C. 112(b); MPEP 2173.05(e)).

I didn't catch this either, but it's a good one. Usually examiners are pretty good about catching antecedent basis issues. I wonder if this indicates more shenanigans. Maybe this is trivial, I don't know. But it's perfectly valid to challenge with a PGR — and not just as another “examiner slipped” gripe: a claim element with no proper antecedent basis is a genuine § 112(b) indefiniteness defect, an actual ground that folds straight into my § 112 case. (Is it worth the effort? Doing the right thing should not have a price tag. The examiner screwed this up; the Office should fix it.)

[Claude Output]
F-2 — Claim typo: claim 4 recites "a previous stoke" for "stroke" (RCE claims, claim 4).

Yep, this made it to the final patent. More typos. It's trivial… but let's look at claim 4 for a second:

[Amended claim & published patent]
4. The sensor system of claim 2, wherein the processor subsystem is configured to determine the direction change of the at least one rod to indicate that a previous stoke has been completed and that a new stroke is beginning.

That's exactly what the whole 3 direction changes thing actually means. It's self-defining. The point is you don't need much from the prior art to see that this is just naked wordsmithing.

[Claude Output]
F-3 — Non-final grouping: claims 19–20 were grouped under "§103 over Phillips in view of Fyfe," yet the body reasons them on Phillips alone with no application of Fyfe (non-final p.11–12) — an imprecise statutory basis later effectively conceded when the Final moved 19–20 to §102/Phillips.

This is where and how Fyfe disappeared! See Dispatch No. 4. If the examiner bothered to read Fyfe at the RCE (or my Third-Party Preissuance Submission) this would have panned out differently, or at least I like to think it would.

---

The contradictory argument

Examiner: Phillips anticipated this.

Applicant: But he didn't state that explicitly the same exact way we said it.

Examiner: Oh, ok.

Why this matters

AI capability and output quality have accelerated sharply over the last several months. The models available in July 2025 were mostly retired.

You guys have no idea how powerful these tools can be in the wrong hands. This story all happened before these tools became widely available. Now it's trivial for anyone to analyze, in depth, an examiner's entire prosecution history — to reverse-engineer what arguments a given examiner does and does not accept. Not whether the thing is actually patentable. Just what works on that particular person. That's not invention, and it's exactly the abuse I'm warning about: my story is about to get put on steroids for everyone else.

To be clear about my own side of this: I've used AI here only to read public documents I'm entitled to read — the file history, the prior art, the patent itself — and to help me put them into plain language. I have not profiled this examiner, and I'm not shaping arguments to anyone's psychology. My PGR stands or falls on the prior art and the law, full stop.

Related dispatches

• Dispatch No. 4 — The first time this idea was good enough to copy — the earlier time this same idea was good enough to copy.
• Dispatch No. 6 — The Third-Party Preissuance Submission — the prior-art submission that should have stopped this.
• Dispatch No. 12 — A Private Conversation, Held in Public — digging through the office actions, I found exactly how this happened.
• Dispatch No. 16 — Claim 13 — Again — the Claim 13 misrepresentation I'm asking about.
• Dispatch No. 20 — New material, not in the disclosure — the undisclosed "previous stroke" material this analysis points to.