Effort Is Owed to the Reader

Source: jason@calacanis.com. "Disgraziad: The WSJ’s Millie Vanilli Editorial Page." August 29, 2026. jcalfromallin.substack.com

The Gist

If you did not spend attention writing it, you are not entitled to the reader's attention reading it.

Conclusion

Undisclosed AI writing is not only deceptive but discourteous, and it invites a symmetrical machine response that dissolves the point of publishing an opinion at all.

Premises

1. Composing one's own thoughts is genuinely difficult, and it is difficult for accomplished writers too, so the difficulty is not a mark of amateurism to be engineered away.
2. Reading attentively is also costly, and the exchange between writer and reader assumes both sides are spending something.
3. Delegating composition removes the writer's cost while continuing to ask for the reader's.
4. An unreciprocated demand of that kind is disrespectful of the reader, which is why the natural response is to have a machine summarize the piece rather than read it.

Assumptions

• Primary sources: Jason Calacanis, All-In Podcast clip posted by @ThisWeeknAI, 29 Aug 2026 (https://x.com/ThisWeeknAI/status/2093770140730957969); and the companion essay "Disgraziad: The WSJ's Millie Vanilli Editorial Page," JCal from All In, 29 Aug 2026 (https://jcalfromallin.substack.com/p/disgraziad-the-wsjs-millie-vanilli). All-In episode chapter "AI writing controversy" at 1:01:33 (https://podcasts.apple.com/us/podcast/all-in-with-chamath-jason-sacks-friedberg/id1502871393). This reconstruction is of Calacanis's published case, not of Druckenmiller's or the WSJ's.
• Steelman of the essay's reciprocity claim. It assumes readers value the writer's expenditure at the keyboard rather than the prior thinking. Druckenmiller's reply (fifteen years on the ideas, composition as the cheap part) is the main counter and is not answered here.
• "Lazy" is treated as a description of the cost structure, not as a character attack.

Analysis

Overall strength: Weak. Argument type: Deductive.

Premise Strength

• Composing one's own thoughts is genuinely difficult, and it is difficult for accomplished writers too, so the difficulty is not a mark of amateurism to be engineered away. (Weak) — Supported only by first-person testimony and an uncited list of famous names (King, Kubrick, Stan Lee) asserted to share this difficulty, not demonstrated. It also silently equates 'difficulty of composition' with 'the effort that matters to readers,' which is precisely what is disputed.
• Reading attentively is also costly, and the exchange between writer and reader assumes both sides are spending something. (Moderate) — A plausible, intuitively appealing model of the writer-reader relationship as reciprocal expenditure, but it is one contestable framing among others (e.g., a model where readers value ideas/accuracy independent of authorial labor) and is asserted rather than established.
• Delegating composition removes the writer's cost while continuing to ask for the reader's. (Weak) — This is the argument's load-bearing premise and its most vulnerable point. It treats 'writer's cost' as located entirely in drafting, ignoring that idea-generation, research, and judgment may remain fully intact under AI-assisted composition — exactly the objection the argument's own assumptions admit is unanswered.
• An unreciprocated demand of that kind is disrespectful of the reader, which is why the natural response is to have a machine summarize the piece rather than read it. (Weak) — The normative claim (disrespect) has some intuitive pull if P2-P3 are granted, but the causal/behavioral claim (readers will naturally retaliate via AI summarization) is an unevidenced prediction, and the prescribed remedy contradicts the very reciprocity norm it is meant to vindicate.

Potential Fallacies

• Unsupported conjunction in the conclusion (Conclusion, relative to P1-P4) — The conclusion asserts the writing is both 'deceptive' and 'discourteous,' but P1–P4 only build a case about cost-asymmetry and disrespect. No premise addresses concealment, misattribution, or reader deception directly, so the 'deceptive' half of the conclusion is asserted rather than derived.
• Unaddressed rebutting defeater treated as resolved (P1/P3, acknowledged in A2 but not integrated into the conclusion) — The argument's own assumptions (A2) concede that the strongest counter — that an opinion piece's real cost lies in years of prior thinking rather than the mechanics of drafting — is 'not answered here.' Yet the conclusion proceeds with full confidence as though the reciprocity violation in P3 is established, rather than treating it as contested.
• Equivocation on 'cost' (P1 through P3) — The argument needs 'cost' to mean the mechanical labor of composition for P3 to establish that delegation removes the writer's expenditure. But P1's intuitive support (the blank page is hard) more naturally describes cognitive/experiential labor — thinking, research, judgment — which AI-assisted drafting may leave fully intact. The two senses are traded on without being distinguished.
• Arbitrary line-drawing (P3, in tension with the source's own carve-outs) — The source material this argument is drawn from explicitly endorses AI delegation for research, brainstorming, proofing, and editing as 'completely valid,' all of which also remove some writer cost. No principled boundary is offered explaining why composition specifically, and not these other delegated tasks, is the ethically decisive site of 'cost' that must not be outsourced.
• Unlicensed strengthening / unsupported prediction (P4 to Conclusion) — P4 asserts that 'the natural response' to perceived disrespect is machine-summarization, presented as a near-inevitability rather than a hypothesis. No behavioral evidence, survey data, or base rate is offered; alternative reader responses (skepticism, disengagement, indifference) are at least as plausible.
• Self-undermining prescription (P4 and Conclusion) — The remedy the conclusion endorses — readers using AI to summarize rather than read — itself abandons the reader's side of the reciprocal expenditure the argument is defending. The 'fix' for broken reciprocity is more withdrawal of effort, which accelerates rather than restores the norm being protected.

Counterarguments

• P1/P3 (High impact) — The strongest and explicitly acknowledged counter (attributed to Druckenmiller): an opinion piece's substantive value and cost lie in years of accumulated thinking and expertise, not in the mechanical act of drafting sentences. If so, AI-assisted composition does not eliminate the writer's genuine expenditure, and the reciprocity violation the argument depends on does not occur.
• P3 (High impact) — Long-standing ghostwriting norms — where a public figure's speech or memoir is substantially composed by someone else and disclosed only via a minor credit — are treated by convention as ethically acceptable, not as inherently disrespectful. This precedent undercuts the claim that undisclosed compositional delegation is uniquely disrespectful, suggesting the real issue is disclosure, not effort-parity.
• Conclusion (High impact) — The prescribed reader response (AI-summarizing suspected AI writing) is not a targeted or proportionate reaction to an individual violator; it degrades attentive reading as a general practice, accelerating exactly the collapse of reciprocal effort and discourse the argument claims to be defending.
• P2 (Medium impact) — Some readers may value only accuracy, insight, or information regardless of who or what produced the prose. If so, the reciprocity model in P2 does not apply universally, and no obligation of 'disrespect' is generated for those readers.
• P4 (Medium impact) — AI-generated text is increasingly difficult to reliably detect, undermining the practical premise that readers can identify when a writer 'stopped spending' and therefore feel the disrespect the argument describes.

Suggested Improvements

• Separate the deception and discourtesy claims — Argue for 'deceptive' and 'discourteous' as distinct claims requiring distinct support — deception via a disclosure/misattribution argument, discourtesy via the reciprocity argument — rather than bundling both into one conclusion supported only by the reciprocity premises. As written, only the discourtesy half is even partially supported by P1-P4; the deception claim needs its own premises about concealment and reader expectations.
• Engage the Druckenmiller counter directly — Explicitly address whether prior intellectual labor (research, judgment, years of thinking) can satisfy the reciprocity requirement even when final composition is delegated, rather than naming this counter in the assumptions and leaving it unanswered. This is identified as the argument's central and self-acknowledged vulnerability; without engaging it, the core premise (P3) remains an assertion rather than an established claim.
• Draw a principled line for acceptable AI use — Provide a non-arbitrary criterion distinguishing condemned AI composition from the endorsed uses (research, brainstorming, proofing, editing), since all of these also reduce writer effort to some degree. Without such a criterion, the boundary between 'legitimate assistance' and 'disrespectful delegation' looks stipulated to match a pre-existing conclusion rather than derived from a consistent principle.
• Support the behavioral prediction in P4 — Offer some evidence or reasoning for why machine-summarization specifically (rather than skepticism, distrust, or simple disengagement) is the expected reader response to perceived disrespect. As it stands, P4's predictive claim is asserted with rhetorical confidence disproportionate to its evidentiary support.
• Address the self-undermining remedy — Acknowledge that prescribing reader-side automation as a response to writer-side automation risks accelerating the very decline in attentive reading the argument is meant to protect, and consider whether disclosure norms alone would better preserve the reciprocal exchange. The proposed 'symmetrical response' currently reads as retaliatory satisfaction rather than a solution consistent with the argument's own stated values.

Scenario Tests

• A writer uses AI only to convert their own fully-developed, original analysis (years of expertise) into polished prose, making no substantive intellectual contribution via the AI beyond phrasing. (Challenges) — If the writer's 'real' cost (thinking) remains intact, P3's claim that delegation removes the writer's cost fails, since only a lower-value mechanical layer was outsourced — directly testing and weakening the argument's central premise.
• A publication adopts mandatory AI-disclosure bylines (as the source material itself proposes), and readers are informed which pieces involved AI composition. (Challenges) — Once disclosure is standard, the 'deceptive' half of the conclusion evaporates, leaving only a much weaker residual claim about labor-based courtesy — suggesting the argument's real force is a disclosure norm, not a blanket prohibition on AI-assisted composition.
• The same reciprocity logic is applied to ghostwritten political speeches or translated literary works, where a named author's composition cost is also removed while full reader attention is still demanded. (Challenges) — Since these practices are widely tolerated with only minor attribution norms, consistent application of P2-P4 would condemn far more accepted practices than the argument intends, suggesting an ad hoc rather than principled line is being drawn around AI specifically.
• A reader who cares only about the accuracy and usefulness of the ideas presented, indifferent to who or what produced the prose. (Challenges) — For such a reader, no 'unreciprocated demand' registers as disrespect, showing that P2's reciprocity model is not a universal description of the reader's stance but one value framework among several.

Coherence & Relevance

The argument has an internally consistent core: if reading and writing are both costly, and delegation removes only one side's cost, then an asymmetry results that can plausibly be called disrespectful. But this core coherence is undercut by an admitted, central, unaddressed objection (that intellectual labor rather than composition is the relevant cost), by an unprincipled boundary between endorsed and condemned forms of AI delegation, and by a conclusion that both overreaches beyond what the premises support (adding 'deceptive' and 'dissolves the point of publishing') and prescribes a remedy that contradicts its own reciprocity standard. The result is a rhetorically effective but structurally fragile argument, best read as an advocacy position in an unresolved dispute rather than a settled demonstration.

• Composing one's own thoughts is genuinely difficult, and it is difficult for accomplished writers too, so the difficulty is not a mark of amateurism to be engineered away. (Moderate) — Functions more as a preemptive rebuttal to an anticipated 'this is just automation of a chore' objection than as a load-bearing premise for the reciprocity conclusion; it does not itself establish that composition-difficulty is what readers value or are owed.
• Reading attentively is also costly, and the exchange between writer and reader assumes both sides are spending something. (Strong) — Sound as a normative framing but not defended against the alternative view that what readers 'pay for' is accurate ideas rather than authorial labor as such.
• Delegating composition removes the writer's cost while continuing to ask for the reader's. (Strong) — This is the pivotal premise, and its major gap is the unresolved conflation between composition-cost and thinking-cost; if prior intellectual labor persists, the premise's factual claim is false in exactly the cases the argument is meant to address.
• An unreciprocated demand of that kind is disrespectful of the reader, which is why the natural response is to have a machine summarize the piece rather than read it. (Moderate) — The normative half (disrespect) follows reasonably if P2-P3 are granted; the descriptive/predictive half (the 'natural response') is a separate, evidentially unsupported claim bolted onto the normative one.

View this argument on LogicFirst.ai