What is a Forward Deployed Engineer, actually?

Job postings for Forward Deployed Engineers are up 729 percent year over year. You have seen that number, and I want to know who counted it.

Here is why I care. This show has a published evidence standard, and it carries a hard rule for industry statistics: every number airs with a named source and an as-of date, and a percentage without a named denominator does not air as fact. Up 729 percent from what? Counted by whom? Counted how? If nobody can answer, the number is a rumor with digits.

Our standard also ranks sources for exactly this kind of claim. At the top sit primary financial disclosures, the filings and the earnings calls. Next come named datasets with published methodology, the government labor statistics and the hiring indexes that state their denominator and their window. Press coverage sits at the bottom, as a pointer to the primary source and never the citation itself. So the question becomes: where on that ladder does the 729 sit?

I went tracing. The trail starts at Business Insider, in May 2026. That is the article everyone is quoting, whether they know it or not. Press, so a pointer. Fine. Follow the pointer. What is the primary source?

The article says Indeed. Specifically, Indeed data shared privately with the reporter. And that is where the trail ends. Indeed did not publish the dataset. There is no methodology page, and there is no report you can download and check. The number everyone quotes rests on data you cannot see.

It gets worse. The posting counts that circulated alongside the percentage are index points, and readers took them for raw counts. An index point is a normalized value, scaled against a baseline period. It measures relative movement. It does not count jobs. An index can jump 729 percent on a tiny base, so without the raw counts you cannot tell a hiring wave from a blip. And the raw counts were never published.

Maybe another tracker has the goods. I checked the competing figures from other trackers, and they carry the same problem: different numbers, and no published methodology behind any of them. So there is no second, independent dataset to triangulate against. Our standard grades a trend claim strong only when multiple independent datasets converge. Here we cannot fully inspect even one.

Let me say the flags out loud. Business Insider: press aggregation, the lowest tier in our hierarchy for jobs-and-money claims. The Indeed data: private, unpublished, no denominator, no methodology. The competing trackers: same gap, different logos.

Here is the verdict, as a grade. The claim that Forward Deployed Engineer postings are up 729 percent year over year is unverifiable. Under our standard it does not air as fact, and the standard says the next part explicitly: when we cannot verify the number everyone keeps quoting, saying so is itself the content. That is this segment. The trend could be real. Hiring for this title may genuinely be surging. We cannot tell, and neither can anyone else repeating that percentage, because the evidence behind it is not public. And to be precise about which claim we are making: this is no evidence, which is different from evidence the trend is fake.

What should you do? Ask two questions every time a viral workforce statistic crosses your feed. Says who, and up from what? If the first answer is a private dataset and the second is an index with no denominator, treat the number as marketing until someone publishes the counts.

And notice what the number was never going to answer anyway: what a Forward Deployed Engineer actually is. The percentage was supposed to prove the role matters, and the percentage fell apart under one afternoon of tracing. The definition is next, and for that question a primary source actually exists. Michael has it.

Sources

• Business Insider, Forward Deployed Engineer postings coverage (May 2026) [Press aggregation]: cites Indeed data shared privately with the reporter; raw counts and methodology never published.
• Indeed postings data (private, unpublished) [index points, not raw counts; no named denominator]: could not be verified.
• Competing tracker figures [no published methodology]: could not be verified.
• The AI Inevitable Evidence Standard, v1.1 (2026-07-04), Industry and discourse claims section: airing rules applied in this segment.

What did Palantir actually mean when it called someone a Forward Deployed Engineer? Whitney just showed you the trend number behind this title falling apart under one afternoon of tracing. The definition survives that treatment, because for this question a primary source exists, and I read it.

Let me flag the source out loud before I use it. The write-ups come from Palantir itself. Palantir is a vendor, and under our evidence standard a vendor grading its own performance stays a claim until someone independent checks it. A vendor describing its own job design sits in a different spot. The company invented the title, staffed the role, and published what the role does, so for the question "what did Palantir mean," Palantir is the primary source. Our standard uses the same logic for price sheets: a vendor's live pricing page is the primary source for its own prices, and for nothing else. So: vendor, primary for its own role, and I will keep every claim inside that boundary. These write-ups establish what Palantir meant. They establish nothing about whether the model outperforms alternatives, because nobody has published that study.

Here is what the write-ups say. Palantir describes two distinct roles that deploy together, and the distinction carries the whole definition. Neither role makes sense without the other, which is exactly the part the industry would later drop.

The first is the forward deployed software engineer. FDSE, in their shorthand. This person is embedded at the customer: physically present, inside the customer's organization, working on the customer's real problems. And this person writes production code. Hold onto that detail, because it carries most of the weight. The FDSE is a software engineer in the full sense, shipping real software, whose desk happens to sit at the deployment.

The second role is the deployment strategist. The strategist is the domain translator: the person who learns the customer's world, figures out which problems are worth engineering time, and translates between the customer's language and the team's. The strategist does the understanding so the engineer can do the building.

Put the two together and you can see what the title named. "Forward deployed" described one half of a paired operating model: an embedded engineer shipping production code, working beside a translator whose entire job is context. Remove either half and you are running a different play under the same flag.

Now the drift. Somewhere between those write-ups and this year's hiring wave, the industry adopted the title. The operating model did not make the trip. Postings borrow the words "forward deployed" while describing work with no embedding, no production code at the customer, and no strategist counterpart anywhere on the org chart. Since no authority owns a job title, every company that borrows it can mean something different by it, and a term with one precise origin turns into a term with no shared meaning.

Here is the verdict, as a grade. On what Palantir meant, the evidence is strong: a primary source, in the vendor's own published words, drawing a clean line between the engineer and the strategist. On what the rest of the industry means by the same three words, no shared definition exists, because none was ever published for adopters to share. Those are two separate claims, and only the first one is settled.

What should you do? When this title crosses your desk, in a posting, a pitch, or a conference bio, check it against the origin. Three markers: embedded at the customer, writing production code against the customer's real problems, paired with a domain translator. Markers present, and the company is running the play Palantir named. Markers absent, and you are looking at a borrowed word, so ask the company what it means. The answer is information either way: a crisp answer means a real model sits behind the word, and a vague one means the word arrived alone. Whitney is next with why titles drift when no method ships with them, and what happened the last time.

Sources

• Palantir, published role write-ups: forward deployed software engineer (FDSE) vs. deployment strategist [Vendor, primary for its own role]: the two-role operating model; embedded engineer writing production code, strategist as domain translator.
• The AI Inevitable Evidence Standard, v1.1 (2026-07-04), Industry and discourse claims section: vendor-primary-for-its-own-facts logic (the price-sheet rule) applied to role definitions.

Should you chase the Forward Deployed Engineer title? Michael just showed you the origin: a precise, published, two-role definition sitting on Palantir's own pages. Before you answer the career question, look at what happened the last time a company coined an engineering title, because the difference between the two cases explains this entire episode.

The last time was Site Reliability Engineering. Google coined SRE, and then Google did the thing that matters for this story: it published a full handbook. Flag the source out loud: the handbook comes from Google, a vendor, primary for its own practice, the same status Michael gave Palantir's write-ups. Now watch what that publication did for everyone downstream. A company that wanted SRE did not have to reverse-engineer the model from job postings. The operating model was written down, in public, at book length, and adopters had a method to copy. The title traveled with its method attached. When another company said it ran SRE, you could pull the handbook and check the practice against the page. The word and the method were one package, so drift had nowhere to start.

Run the same test on Forward Deployed Engineer. Palantir published role write-ups; Michael walked you through them, and as a definition they are strong. What Palantir did not publish is the method. No handbook. No equivalent of the book that taught the industry SRE. So companies adopting the title had nothing to copy except the word itself, and the word is exactly what they copied. Google shipped a title with a manual. Palantir shipped a title without one, and every adopter filled the blank with whatever it wanted the words to mean.

That asymmetry is the mechanism behind everything in this episode: a viral number nobody can check, attached to a title nobody defines the same way, orbiting a role that has a real, precise definition at exactly one company. Nothing in that chain requires bad faith. A vacuum where the method should be gets filled, and it gets filled fast when a title is hot.

So, the career question. If you are deciding whether to chase this title, here is how our standard translates into a job search. First, judge the role by its written responsibilities, because the three words at the top of the posting carry no shared meaning. Read what the job has you do all day, and if that matches Michael's markers, embedded at the customer, production code, a translator counterpart, then it is the real play regardless of what the header says. Second, ask any employer using the title what they mean by it. A company running the model can answer in a sentence. A company decorating an ordinary role with a hot title will stumble, and the stumble is your data. Third, treat compensation rumors without methodology as unverified. Pay numbers ride this trend the same way the posting numbers do, and they arrive with the same missing denominators.

Here is the episode verdict, as grades. A real role exists at its origin: strong, on Palantir's own published write-ups, vendor but primary for its own role. An unverifiable trend number rides on top of it: the 729 percent figure I traced in the first segment never produced a public dataset, so it does not air as fact on this show. And what the industry means by the title has no shared answer, because no method was ever published for adopters to copy. Title adoption without method: that is the FDE phenomenon.

Now the practical close for the whole episode. When the next hot title crosses your feed, ask who counted the trend and up from what. Then look for an origin definition, the way Michael found one here. And if you are interviewing, judge the responsibilities on the page, ask the employer what the title means to them, and hold every number attached to the trend to the same bar we used in these three segments.

Sources

• Google, Site Reliability Engineering handbook [Vendor, primary for its own practice]: the published method SRE adopters could copy.
• Palantir, published role write-ups (see V2 sources) [Vendor, primary for its own role]: no equivalent published method for FDE.
• Business Insider via private Indeed data (see V1 sources) [Press aggregation; unpublished dataset]: the 729% claim, unverifiable per the V1 trace.
• The AI Inevitable Evidence Standard, v1.1 (2026-07-04), Industry and discourse claims section: airing rules and grading applied to the episode verdict.