Sometime around 2015, hundreds of technology companies began reorganizing themselves into Squads, Tribes, Chapters, and Guilds. The vocabulary came from a blog post about how Spotify had structured its engineering organization. The model was coherent, attractive, and provided enormous quantities of status to whoever introduced it. There was one problem: Spotify had already moved on from it. The company had found that the model didn't work for them. The blog post's author said this repeatedly, at conferences, in interviews, on the internet. It didn't matter: the language game had escaped its origin and was now self-sustaining, propagating through the industry on the strength of its own vocabulary rather than any demonstrated results.
Wittgenstein's Philosophical Investigations introduced Sprachspiele to show that the meaning of a word is not some inner essence but simply its use within a particular social practice. Different games are not just different vocabularies for the same underlying reality — they constitute different realities. Drop a user into a SAFe PI Planning event and you have your lion who speaks but cannot be understood.
The language games of software development (Agile, Scrum, OKRs, "digital transformation,") have now subsumed value delivery as the reason of work. The game comes to feel like the thing itself, and this abstraction serves specific interests at the expense of the one party who is never in the room: the person the software is supposed to help.
The Agile Manifesto of 2001 was a genuinely game-changing document. Seventeen practitioners gathered at a ski resort and produced a text that redefined how people thought about software development. "Working software over comprehensive documentation." The thing they were rebelling against was waterfall project management, which had imported its vocabulary from manufacturing and applied it to knowledge work with the confidence of someone using a power drill to do surgery. The problem, as with all successful rebellions, was institutionalization. Within a decade, "Agile" had become a multi-billion dollar certification industry. SAFe was essentially Taylorist management reorganized into Agile vocabulary — PIs instead of quarters, ARTs instead of departments. Ron Jeffries, one of the Manifesto's original signatories, eventually published "Developers Should Abandon Agile." Dave Thomas declared "Agile is Dead" at a 2014 keynote. Neither intervention made much difference, because as Kuhn observed in The Structure of Scientific Revolutions, paradigm shifts are generational. The game perpetuates itself through credentials, hiring patterns, and software tools.
Atlassian's Jira is the perfect artifact here. Jira literally encodes the language game into the interface. Your work doesn't exist until it exists as a ticket. The software enforces the categories of the game on anyone who touches it, making the game appear natural, inevitable, the way things simply are. To ask "but does this ticket actually matter to anyone?" is to ask a question that Jira provides no field for.
Harry Frankfurt, in On Bullshit, distinguishes between the liar, who knows the truth and asserts the opposite, and the bullshitter, who is simply indifferent to truth. "Digital transformation" is bullshit in Frankfurt's technical sense — not false, exactly, but indifferent to its own truth value. Gartner, McKinsey, and Deloitte all define it differently; none of those definitions constrain what a client can bill for. Watch how the vocabulary propagates through a strategy engagement. "North Star metric" enters to imply that a single number can represent organizational purpose. "Operating model" imports manufacturing language to suggest human judgment can be treated as a variable to be tuned. "Value stream" implies software development flows like liquid through pipes, ignoring that knowledge work is primarily about discovery and judgment and the kind of tacit intelligence that cannot be pipelined. These forms perform certainty; they do not produce it. The standup meeting looks like information sharing. The roadmap looks like a plan. Both are often something else entirely.
James Scott's Seeing Like a State distinguishes between techne — formal, codified technique — and metis — practical wisdom, the kind of knowledge that cannot be written down because it is constituted by experience and context. Scott's argument is that modern institutions consistently attempt to replace metis with techne, because techne is legible to central authorities and metis is not. You can audit a checklist. You cannot audit wisdom. Applied to software development: a senior developer's tacit understanding of where the user actually struggles, accumulated through support tickets and hallway conversations and the way someone hesitates before clicking a button — all of this is illegible to a velocity dashboard. Converting it to story points makes it manageable. It also makes it wrong. And Goodhart's Law guarantees things will diverge from there: when a measure becomes a target, it ceases to be a good measure. Sprint velocity, NPS scores, DAU, conversion rates — all of them are gamed, not usually through explicit dishonesty, but through the natural tendency of people to optimize for what is measured at the expense of what is not.
The user doesn't disappear from the language game. They appear constantly. "User story." "User-centered design." "Customer obsession." The user is everywhere in the vocabulary and nowhere in the room. What you actually find in sprint planning are personas — fictional composites named things like "Sarah, 34, project manager" and given backstories designed to make empathy feel available without requiring its actual exercise. The persona is a user-shaped object that fits comfortably into a PowerPoint deck and never contradicts you. Clay Christensen's Jobs to Be Done framework is one of the more serious attempts to cut through this. Stop asking what features users want and start asking what progress this person is trying to make in their life. A Scrum team might have a user story about faster report loading — a clean, well-formed story that will look good on the sprint board. The JTBD question is: what job is this person trying to do? Often the answer is "make a decision," and the report is a workaround for a broken communication process. The real problem is organizational, and no sprint velocity metric would surface it. The language game has pre-structured the problem so that the answer will always be some form of "build more features."
Simone Weil's concept of attention, developed in Waiting for God, names what is actually required. Genuine attention is not active imposition of our frameworks on reality. Instead, it is a kind of self-emptying; letting the impulse reveal itself. Contextual inquiry, where you sit with a user in their actual context watching them do their actual work, operationalizes something like Weil's attention. It is widely acknowledged and rarely practiced at scale. The language games are institutionally preferred to methods that would destabilize them, and that preference is not accidental.
Some people in every Agile organization know exactly what they are doing. Consultants who sell SAFe transformations know that "transformation" is doing rhetorical work rather than descriptive work. Vendors who build Jira know that encoding a methodology in software creates switching costs that sustain the methodology's dominance regardless of its results. Bourdieu's concept of symbolic capital explains why the games are sticky: mastery of the vocabulary confers status independently of whether the vocabulary tracks reality. The Scrum Master role, the Agile coach position, the transformation consulting practice are economic interests dressed in methodology vocabulary. This doesn't make the individuals cynical; most of them genuinely believe in what they're doing. The abstraction between effort and value is the problem.
The Agile Manifesto is comprised of four values and twelve principles, which fits on a single page. In contrast, a description of a SAFe PI Planning event: the two-day ceremony involving an ART, multiple teams, Business Context, Product/Solution Vision, Architecture Vision, Team Breakouts, Risk Management, and Confidence Votes. The distance between these two things is the subject of this essay.
Wittgenstein thought philosophy's task was therapeutic: we cannot use philosophy to prescribe and build better systems, but to dissolve the confusions that language creates. The therapeutic move is noticing when language is bewitching you, when you are mistaking the ritual for the accomplishment. What would you do if you couldn't use the word "sprint"? What would you do if "velocity" was unavailable? What would you do if the user story template was gone and you had to describe, in your own words, what you watched someone struggle with last Tuesday?
The honest answer, in most organizations, is: nothing. The language games are not obstacles to work. They have become the work. And that is what makes them so difficult to escape, and so worth trying.