The Illusion of Output: Process Goals and Real Capability

The Illusion of Output: What a Result Cannot Tell You About the System Behind It

An outcome can be real and still tell you almost nothing about the system behind it.

A team can hit the number by borrowing from next quarter. A person can receive a title without earning the trust needed to lead. A project can look finished because the hardest judgment was supplied by someone else. The output exists. The capacity to reproduce it may not.

Episode 115 began as a broad argument about shortcuts, health, status, leadership, AI, and the growing ability to acquire the visible result without following the path that used to produce it. The evidence review narrowed the public version to a more useful question:

What remains after the outcome?

If the answer is a repeatable process, better judgment, usable evidence, and the ability to correct the next attempt, the result built capacity. If the answer is only the result, the system may still be missing.

flowchart LR
    A["Desired outcome"] --> B["Controllable process"]
    B --> C["Performance evidence"]
    C --> D["Feedback and correction"]
    D --> B
    C --> E["External result"]
    E --> F{"Can the system produce it again?"}
    F -->|Yes| G["Capability remains"]
    F -->|No| H["Output without a durable foundation"]

An outcome becomes durable when process, evidence, and correction can reproduce it.

The podcast returned through a small process

When I started publishing Venture Step consistently again, I did not set a download target or require a guest every week.

The minimum commitment was to record for at least ten minutes each week.

That sounds small beside the ambition of building a serious show. It was useful because it was smaller. I could complete it during a weak week, without a guest, before the topic felt perfect. The process kept creating raw material, practice, and evidence.

The outcome still mattered. I wanted a strong show that reached the right people. The minimum weekly repetition kept that outcome from becoming a fantasy with no scheduled behavior beneath it.

The ten-minute floor did not produce the entire editorial system. Recording alone does not verify a fact, preserve a transcript, develop a public page, find a source, or make an article worth reading. The repetition created enough material to expose the next constraints.

That is how a process should grow. It earns another layer when evidence shows what the current layer cannot do.

Three goal types serve one system

Episode 115 used three goal categories.

An outcome goal describes the external result, such as winning. A performance goal defines a personal standard, such as running a target time. A process goal directs attention toward a behavior or strategy, such as executing a planned pace or completing a specific training action.

The categories are not rivals. The outcome chooses direction. The performance measure provides evidence. The process tells the person what to do next.

For Venture Step, the desired outcome was a trusted and useful show. A performance standard could include publication reliability and editorial quality. The process was the weekly recording commitment.

The process is deliberately less impressive than the outcome. It has to survive the week when motivation, timing, guests, or confidence do not cooperate.

[[Process Goals vs Performance Goals vs Outcome Goals]] develops the full goal stack and the failure mode of each layer.

The evidence supported the direction, not the headline

The episode cited Williamson and colleagues' systematic review and meta-analysis of goal setting in sport. The authors screened 17,841 records and included 27 studies.

They reported pooled performance effects of 1.36 for process goals, 0.44 for performance goals, and 0.09 for outcome goals.

During the recording, I divided 1.36 by 0.09 and described process goals as having a fifteen-times advantage. The episode title preserves that original framing.

The literal multiplier was wrong.

The Cochrane Handbook explains that a standardized mean difference is expressed relative to variability and reported in standard-deviation units. It is not a ratio measure. Dividing two standardized effects does not tell us how many times more effective one intervention is.

The paper supports a bounded and still useful conclusion: process goals had the largest pooled performance effect in the included sport literature. It does not establish that they are fifteen times better in sport, business, health, or life.

[[Does Research Really Say Process Goals Work 15 Times Better]] preserves the complete correction so the attractive statistic does not keep traveling after its interpretation failed.

Another number needed correction

The recording also referred to 38 percent of labor being wasted.

The underlying World Economic Forum release said its 2016 Human Capital Index found that 65 percent of global human-capital potential was being optimized through education, skills development, and deployment. The corresponding gap was 35 percent.

That is not the same claim.

The WEF index covered 130 countries and examined how human capital was developed and deployed across people's lives. It did not measure all labor and declare a share of it wasted.

This correction matters beyond one old number. Output-oriented storytelling rewards a crisp statistic. Evidence requires preserving what the source actually measured.

A source can be real, the number can appear on the page, and the public claim can still be wrong because the words around it changed the meaning.

Process is not a religion

The correction does not require replacing outcome worship with process worship.

A process can become another illusion of output. The person completes the workout, publishes the post, sends the messages, attends the meeting, or fills the tracker. The activity is visible. The underlying performance does not improve.

Repetition without feedback can make the wrong behavior durable.

The outcome chooses which processes deserve attention. Performance evidence determines whether the current method is working. The process earns repetition only while it contributes to learning or progress.

This is why the ten-minute podcast rule was a floor rather than a definition of a good episode. It protected continuity. It did not excuse weak research, thin writing, or uncorrected claims.

A plan should start before motivation arrives

A process becomes more reliable when the cue and response are chosen in advance.

Peter Gollwitzer and Paschal Sheeran's implementation-intentions review describes if-then plans that link a specific situation with a goal-directed response. The underlying meta-analysis synthesized 94 tests and reported a medium-to-large overall effect on goal attainment.

"I will work on the show weekly" leaves the start open. "If the Thursday recording window begins, I will open the setup and complete the weekly minimum" chooses the cue and response before the moment becomes negotiable.

That planning evidence should not be turned into another slogan. A precise schedule cannot make a poor action useful. It helps translate an existing goal into behavior.

The process still needs evidence and review.

Friction can build capability or attachment

Episode 115 also argued that removing every difficult step can remove part of the learning.

That idea needs a careful boundary.

Robert and Elizabeth Bjork's work on desirable difficulties distinguishes visible performance during practice from longer-term retention and transfer. Conditions such as spacing, variation, generation, and retrieval can make practice feel slower while supporting later learning.

The authors emphasize that many difficulties are simply undesirable. A challenge becomes useful when it triggers productive encoding or retrieval and the learner has enough knowledge or skill to respond.

The original IKEA-effect research adds the warning. Participants placed higher value on products they had successfully assembled. Labor changed valuation. It did not establish objective quality.

Effort can build skill and make us overvalue the thing we made. A difficult process needs external evidence, not moral protection.

[[Friction Is Not the Enemy of Expertise]] turns that tension into a decision test for what to automate and what learning function to preserve.

AI can produce the result before the operator has the judgment

AI makes the difference between output and capability unusually visible.

A person can now produce plausible copy, code, analysis, images, or plans before they could have created the same output alone. This can expand participation, remove clerical work, and let experienced people spend more time on the decisions that matter.

The risk is not that the assisted result is automatically fake. The risk is treating the result as proof that the operator can detect an invented source, see a broken assumption, handle a novel exception, or explain why the recommendation applies.

The answer is not to make people perform historical busywork as a test of authenticity.

Automate the search, setup, normalization, calculation, and formatting that teach nothing. Preserve accountable review where the person still owns the consequence. Test whether the combined person-and-tool system produces a strong result, and separately test whether the human can recognize the failures the role requires.

The capability belongs wherever the system will need it during the exception.

The medical material does not belong in the public lesson

The original episode opened with GLP-1 medications and made broad statements about hormones, body composition, nutrition, treatment, and shortcuts.

That material is held from the public package.

Medication belongs inside current clinical evidence and a qualified relationship between patient and clinician. It should not be used as a morality play about whether someone "earned" an outcome. A health result can involve treatment, behavior, biology, access, risk, and individual circumstances that this episode did not review adequately.

Removing that section strengthens the public argument. The difference between output and capacity can be explained without turning another person's treatment into a symbol.

The Liver King and Elizabeth Holmes examples are also unnecessary here. Their public stories do not establish the goal-setting evidence, and amateur personality analysis would distract from the operating question.

A durable outcome leaves a system behind

The deepest form of the illusion of output is a result that disappears as soon as the tool, title, temporary pressure, or favorable condition is removed.

A durable result leaves a process, a skill, a relationship, a body of evidence, a better model, or a clearer judgment behind it. The person or organization can explain what happened, recognize the important conditions, and perform again.

That does not make every shortcut bad. Tools should remove work that teaches nothing. Assistance should make valuable capability more accessible. A correct one-time result can still be worth having.

The question is what the next situation will require.

If the person must reproduce, adapt, diagnose, supervise, teach, or recover the work, the system needs more than the output. It needs evidence that the required capability exists in the person, tool, process, or combination that will face the next case.

Use the outcome for direction. Use performance evidence for truth. Use the process for action. Then let the evidence change the process.

Episode 100 offers a related founder operating-system view. Episode 116 shows how an operating system replaces fragile personal coordination. Episode 119 extends the argument into testing under real conditions.

Sources

The first-person podcast process and original claims come from [[E115 Full Transcript]]. The goal-setting evidence comes from Williamson and colleagues' systematic review and meta-analysis. The statistical boundary comes from the Cochrane Handbook. The corrected human-capital statement comes from the World Economic Forum. Cue-linked planning comes from Gollwitzer and Sheeran. The friction boundary comes from the Bjorks and the IKEA-effect paper.

AI assisted with organization, source comparison, and editorial review. Dalton Anderson's transcript and the linked primary sources control the factual claims.