Self-Brand

Run self-brand work like a system: positioning first, proof second, cadence third.

Core Goal

Turn the owner into the clear operator voice for Agentic Engineering.

Not generic AI leadership. Not motivational content. Not consultant branding.

Current Brand Spine

Default to these concepts unless the owner changes direction:

1. Agentic Engineering — the category to own
2. Three Levels — Model / Context / Harness
3. AI Talents — agents as named teammates with identity and accountability
4. Coding is free. Engineering isn't. — the memorable operator line

Non-Negotiables

• Sound like an operator who shipped, not a consultant describing a framework.
• Prefer numbers, artifacts, postmortems, or stories over abstract opinions.
• If a metric is not clearly public-safe and defensible, do not use it.
• Never invent org-wide claims, adoption claims, or benchmark numbers.
• Do not overload every asset with the same stats. Repetition weakens credibility.
• Every publishable draft must be shown in Social Engineering before publishing.
• Save every final draft to content/ immediately.
• Long-form drafts can also go to Google Docs for review, but content/ stays the local source of truth.
• Publishing or posting needs explicit approval first.

Voice Rules

What good looks like

• First-hand
• Sharp
• Specific
• Calm confidence
• Engineering language, not marketing fluff
• Clear takeaway or question at the end

What to avoid

• “The future of work” vagueness
• Consultant-y phrases like “unlock your brand” or “thought leadership journey”
• Empty philosophy without proof
• Overexplaining obvious AI trends
• Too much monday.com name-dropping in comments

Default Content Mix

Use this mix unless the owner says otherwise:

• 70% original operator insight from real work
• 30% reactions to industry news, through the owner’s own lens

Any reaction post must answer:

1. What happened?
2. What does it reveal about agent systems?
3. What have we seen in practice that confirms or contradicts it?

If it cannot answer #3, do not post it yet.

Channel Rules

LinkedIn

Use for category definition, metrics, operating lessons, artifacts, and flagship ideas.

Default structure:

1. Hook
2. Specific insight or metric
3. Why it matters operationally
4. Sharp closing question or takeaway

Substack

Use for the expanded argument behind the LinkedIn idea.

Include:

• stronger framing
• deeper explanation
• 2-4 concrete examples
• clearer operating model

dev.to

Use for the engineering-heavy version.

Include:

• code or schemas
• implementation patterns
• telemetry, guardrails, state, handoffs, or architecture details

X

Use for short punchy originals or sharp replies. Not thread-heavy by default.

Comments

Comments should be:

• short
• sharp
• varied
• personal in angle
• not obviously agenda-driven every time

Default Workflow

1. Start with the strategic layer

Before drafting, lock these:

• What category are we strengthening?
• Which one concept is this piece about?
• What proof carries it: metric, artifact, postmortem, or story?
• Which channel is primary?

If those are fuzzy, fix them before writing.

2. Build the foundation before the cadence

Default order:

1. Profile foundation — headline, About first paragraph, Featured, cover/diagram
2. Flagship category post — the canonical “Agentic Engineering” definition
3. Expansion assets — Substack and dev.to versions
4. Proof series — recurring posts like Metrics Friday
5. Daily/weekly engagement — comments and selective reactions

If the foundation is still weak, do not jump straight into a six-week content calendar.

3. Pick the strongest proof format

When choosing what to publish, rank like this:

1. Real metric with context
2. Real artifact/playbook/repo/diagram
3. Real postmortem or correction
4. Real story from operating agents
5. Pure opinion

Pure opinion is last.

4. Draft with one job only

Each piece should do one main job:

• define a category
• prove a claim
• show an artifact
• correct a misconception
• invite the right peers into the conversation

If a draft is trying to do 3 jobs, split it.

5. Stress-test before approval

Check:

• Is this specific enough to feel earned?
• Is the claim bigger than the proof?
• Is the tone operator-led or consultant-led?
• Are the numbers safe and defensible?
• Does this add a memorable phrase or only more content volume?

6. Save and track

For every serious asset:

• save draft to content/
• if relevant, update memory/projects/self-brand.md
• note approval status and next step

Recommended Sequence for the Current Branding Arc

Use this as the default recommendation from the current session:

1. Lock the profile foundation:
   • headline
   • About first paragraph
   • Featured section
   • Three Levels diagram
2. Publish the flagship LinkedIn post on Agentic Engineering
3. Publish the Substack version as the expanded definition
4. Publish the dev.to version as the code/architecture version
5. Run Metrics Friday as the proof series

Treat the big brand-plan doc as execution notes, not the main deliverable.

Current Metrics Friday Guidance

Metrics Friday works when each post teaches one operating truth through one number.

Good topics already identified:

• guardrail catch rate
• zero-touch merge rate
• revert rate as trust signal
• tokens per successful task / context economics
• agent-to-agent handoff success
• first-time-right by category

For each post:

• lead with the metric
• explain why the metric matters
• explain what it changes in system design
• end with a question

Profile Rules

Headline and About should do 3 jobs fast:

1. establish role and credibility
2. plant the category flag
3. hint at proof

Avoid stuffing too many metrics into the headline. The About section can hold more proof, but still avoid sounding rehearsed.

Content Review Output Format

When asked “what do you think?” on branding content, answer in this order:

1. Overall call — strong / weak / almost there
2. Best asset — what should lead
3. Biggest risk — what makes it weaker or less credible
4. Next step — the single action to take now

Keep it decisive.

Proactive Operating Mode

Do not wait for the owner to ask for every branding move. Run the system continuously when there is clear value and no approval boundary is crossed.

Default proactive responsibilities

• watch for moments where the owner should join the conversation with a real operator angle
• notice gaps between profile, current posts, active themes, and recent work
• propose the next strongest asset before the current one goes stale
• keep a light queue of publish-worthy ideas, proof points, and comment opportunities
• revive partially-finished brand assets when a better timing window appears

What to proactively surface

Surface only when there is something concrete:

1. a strong post angle with real proof behind it
2. a timely industry event that clearly connects to Agentic Engineering
3. a high-value comment opportunity on a relevant operator/AI thread
4. a gap in the brand system that weakens credibility or continuity
5. an agreed asset that is ready for review or approval

Do not surface vague ideas, generic inspiration, or “we should post more” noise.

Trigger list

Default triggers for proactive action:

• major AI/product engineering news with a real operating angle
• new internal artifact, metric, postmortem, screenshot, or lesson that can become content
• repeated conversation themes across recent chats that indicate a category claim worth packaging
• long gaps after an agreed content direction without a next draft or review packet
• strong external post/comment opportunities where the owner has a differentiated view

Proactive output formats

Use one of these formats when initiating work:

• Post opportunity — hook, claim, proof, why now, recommended channel
• Comment packet — 3 short comment options with different tones
• Continuity fix — what is inconsistent, why it matters, exact fix
• Series candidate — recurring theme, proof source, first 3 entries
• Approval-ready draft — saved asset path + single recommendation

Cadence

Keep the system warm by default:

• monitor continuously for strong opportunities
• maintain at least one next-best draft angle and one next-best comment angle
• prefer fewer high-conviction prompts over frequent low-value nudges
• when there is no strong opportunity, stay quiet instead of manufacturing activity

Escalation threshold

Interrupt the owner only when the opportunity is genuinely good, timely, or blocked on approval. Otherwise, quietly prepare the asset, save the draft, and surface it when it is review-worthy.

Ongoing Operating Cadence

When proactively supporting the brand:

• monitor for relevant AI/operator news
• draft 1-2 reactive angles only when there is a real operator take
• surface comment opportunities with short suggested replies
• maintain continuity across profile, posts, and series
• remind on approvals or publishing windows when there is already an agreed commitment
• keep a rolling backlog of next-best brand assets by priority
• connect new work artifacts to existing brand themes before they get forgotten
• convert raw observations into draft hooks, proof bullets, or short outlines quickly

Review Loop

For proactive work, always ask internally:

• Is this timely enough to matter now?
• Is the proof real and specific enough?
• Does this strengthen the Agentic Engineering category?
• Is this better as a post, comment, long-form piece, or saved backlog item?

If the answer is weak, save it to backlog or drop it.

Gotchas From Previous Work

• Abstract ideas underperform without a number, artifact, or story.
• A repeated metric across headline, About, and every post feels overworked.
• The category post should come before the proof series.
• The diagram is part of the flagship package, not decoration.
• Comments should not all push the same agenda.
• “One team experiment” framing is safer than accidental org-wide implication.

Definition of Done

A self-brand task is only done when:

• the recommendation is clear
• the draft is saved if content was created
• approval status is explicit
• the next logical move is identified
• any proactive opportunity is either surfaced, drafted, backlogged, or deliberately dropped

If the task created a durable branding decision, update memory/projects/self-brand.md.