For most of the year you have been the strategist inside the company — running the program, calibrating the eval set, defending the budget, drafting the board update, mapping the 2027 roadmap. The work has been internal. The audience has been your CTO, your CFO, your team, your board. The proof has been your dashboards. The reward has been the runway to do the next quarter of building. This final lesson asks a different question. When does the strategist's voice need to leave the building? When does the company benefit from the strategist being publicly visible — giving a talk, publishing a post, contributing to open source — as the named human behind a real, working agent program? And when the answer is "now," what are the right venues, the right cadences, and the right alignment with the company's communications goals so that the public voice strengthens the program instead of becoming a liability? Three venues stand out for the agent strategist in mid-2026: the AI Engineer Summit (Latent Space's twice-yearly gathering), the Agent Skills GitHub repository (Anthropic's public collection of reusable agent skills), and the Langfuse open-source project (the de facto OSS observability layer for agent programs). This is the final lesson of the program. The arc returns to where it started, but on the other side of a year of building.

Why the Strategist Needs a Public Voice

For most of the program, the case for a public voice has been deferred. There was a program to build, a function to design, a budget to defend. Now the program is real. The function exists. The budget is established. The work is producing results. At this stage, three forces start to compound, and the strategist who does not engage them leaves value on the table.

Force one: hiring leverage

The strategist's company is hiring. The L3 and L4 candidates who would make the program meaningfully better are reading. They are scanning conference speaker lists, GitHub commit histories, technical blog posts, and Substack newsletters before they accept a recruiter's first call. A program with a public voice attracts candidates who otherwise would not respond. A program without one is invisible to the top of the market.

The 2026 hiring math is unforgiving. A senior agent engineer offer at $250K base receives twenty inbound recruiter messages per week. The candidate's question is not "is this a real job" but "is this a place where the work would be technically interesting and the program would be well-run." A talk at AI Engineer Summit, a tracked commit to a public agent skill, or a thoughtful technical post answers that question without the candidate ever having to ask it.

Force two: customer trust

Enterprise customers buying an AI-powered product in 2026 are looking for evidence that the vendor understands what it is selling. They are tired of glossy claims. They are reassured by visible technical depth from the team building the system. A strategist who has presented the company's approach to evals at a public conference is providing exactly the kind of evidence the customer's procurement and security teams want to see.

Customers do not need to attend the talk. They Google the strategist before the renewal call. The slide deck on the strategist's blog, the GitHub repo with the strategist's name on commits, the conference talk on YouTube — these are the social proofs that move a marginal renewal into a clean one and a marginal upsell into a signed expansion.

Force three: market signal

The agent industry in 2026 is still finding its center of gravity. Conventions are being established, terms are being defined, standards are being adopted. The companies whose strategists participate in those conversations get to shape them. The companies whose strategists do not get to live with whatever conventions the others settled on. Influence is not vanity; it is the substrate on which the company's roadmap is going to land.

The strategist's public voice is not a brand exercise; it is an inbound lead generator for hires, an evidence file for customers, and a seat at the table where the industry conventions are written. Companies that fund their strategist's external presence get a return that does not show up in the quarterly P&L but shows up in the hiring funnel, the customer retention rate, and the speed at which the company can adopt new standards.

Alignment With the Company's Communications Goals

The strategist's public voice does not exist in isolation. The company has a marketing team, a comms team, a legal team, and a CEO or founder whose voice is already in the market. The strategist's external work either aligns with those existing voices and amplifies them, or it contradicts them and creates a problem the strategist will eventually have to manage. The first conversation in this work is not with the conference organizer; it is with the head of Marketing or Comms.

The three-question alignment conversation

The strategist sits down with the head of Marketing for thirty minutes and asks three questions:

• What story is the company telling externally this year? Is it a category leadership story, a product launch story, an industry-vertical story, a thought-leadership story on a specific topic? The strategist's work should ladder up to that story, not create a parallel one.
• What topics are off-limits or sensitive? Customer names without explicit permission. Specific revenue numbers. Competitive comparisons that legal has flagged. Vendor relationships under NDA. Internal hiring details. The strategist's work needs the same fences the company's other voices already respect.
• Who reviews before publication? Most companies have a lightweight review process for external content from named employees. Conference talks: comms gets the deck twenty-four hours before. Blog posts: comms gets a draft seventy-two hours before. Open-source contributions: legal gets a one-line FYI if the contribution touches IP. The strategist who learns the process and follows it has a long runway; the strategist who skips the process has a single incident that ends the runway.

The Marketing head will almost certainly say yes. They will probably be enthusiastic. A technical leader willing to do external work is rare; a technical leader willing to do external work that aligns with the company's positioning is invaluable. The Marketing head may even offer to fund the travel, produce the slide template, ghostwrite the first draft of the post, or introduce the strategist to a conference organizer. Take the help. The work is more impactful when it is collaborative.

What contradicts the company story

A strategist whose company is pitching a "your data never leaves your environment" story should not give a conference talk on "how we use customer data to fine-tune our models." A strategist whose company is positioning as a regulated-industries vendor should not blog about "how to ship agents fast and break things." A strategist whose CEO just said publicly "we built our agent platform in-house" should not present a talk titled "why we standardized on a third-party agent platform." The strategist's external voice must be a consistent extension of the company's voice, not a contradiction that creates a Twitter dust-up the comms team has to clean up.

When the strategist genuinely disagrees with the company's external positioning, the right move is to surface the disagreement internally and either change the positioning or abstain from the external work. The wrong move is to publish the disagreement and learn afterward that it created a problem. The strategist who has earned the right to public visibility has also earned the responsibility to align it.

Venue One: The AI Engineer Summit

The AI Engineer Summit, organized by Latent Space (swyx and the team), runs twice a year — typically World's Fair in San Francisco in June and a follow-up event in late fall in New York. The audience is roughly 2,500-3,500 in-person attendees plus 30,000-50,000 livestream viewers, weighted heavily toward technical staff at AI-native and AI-adopting companies. The talks are practical, technical, and pattern-rich. The bar for selection is reasonably high but achievable for a strategist who can talk about a real production system with real numbers.

What a strategist's talk looks like

A good AI Engineer Summit talk from an agent strategist is not a vendor pitch and not a futurism keynote. It is a 20-30 minute walkthrough of a specific problem the strategist's team solved, with the architecture diagram, the eval numbers, the failure modes, the costs, and the lessons. Titles that have worked in 2025 and 2026:

• "How we cut our customer-onboarding agent's tail-latency from 47 seconds to 8 by changing how we cache tool calls." A specific problem, a specific solution, transferable lessons.
• "Trace review as the operating discipline: how we run a weekly meeting that has caught 12 production regressions in six months." A ritual, an outcome, a methodology others can copy.
• "From pilot to production: the eval set that let us ship a financial-services agent past compliance review." A regulated workload, a specific bar, a reproducible approach.
• "The cross-functional agent pod: how we structured a six-person team that ships an agent every two months." An org-design pattern, a cadence, the rationale.

The format that consistently works: 5 minutes of context (what we were trying to do, why), 15-20 minutes of the technical detail (what we built, what went wrong, what we changed, what worked), 5 minutes of takeaways (the patterns the audience can apply). One architecture diagram. Two or three real numbers. One honest failure that the team learned from. No vendor logos beyond the ones the team actually used. No "future of work" speculation.

How to get accepted

The AI Engineer Summit reviews speaker proposals through a public CFP, opens roughly four months before the event, and the selection committee favors specific over abstract, technical over strategic, and produced over polished. The strategist with a clear topic, a one-paragraph abstract that names the specific system and the specific outcome, and a link to one prior public artifact (a blog post, a small open-source contribution, a meetup talk recording) is in the consideration set. The strategist with a generic "the future of AI agents" abstract is not.

The hack that works: write the blog post first, then submit the talk based on the blog post. The post becomes the proof of substance, the talk becomes the amplification. The selection committee can see the depth before they accept the talk.

What the company gets

An AI Engineer Summit talk by the strategist produces three durable assets. The 25-minute YouTube video that the company's recruiters send to every senior candidate for the next two years. The slide deck that the company's sales team excerpts for procurement conversations. The introductions the strategist makes in the hallway track to peer strategists at other companies, who become reference customers for the company's product and resource exchanges for the strategist's team. None of these are billable in the quarter; all of them are durable.

The AI Engineer Summit is the conference the agent industry's senior practitioners attend. The strategist who presents there is meeting the L4 and L5 candidates the company wants to hire, the customer architects the company wants to renew, and the peers whose mutual respect makes the next industry standard easier to ship. Forty minutes of stage time produces twelve months of compounding return.

Venue Two: The Agent Skills GitHub Repo

The Agent Skills repository (github.com/anthropics/skills and adjacent community forks) is Anthropic's public collection of reusable agent skills — packaged prompts, tool definitions, and evaluation harnesses that other developers can clone, adapt, and ship in their own agents. Through 2025 and 2026 the repo has become the de facto reference library for "how to do X with an agent": how to triage support tickets, how to do financial-document extraction, how to run a multi-step research workflow. Contributions are visible. Quality is curated. Maintainers (Anthropic staff plus a growing community of high-signal contributors) merge well-thought-out additions and reject thin ones.

What a contribution looks like

A good contribution to the Agent Skills repo is not a one-off prompt. It is a self-contained skill: a clear problem statement, the prompt or set of prompts, the tool definitions (typically MCP-compatible), an eval set with at least 30-50 cases, a README that explains when to use the skill and what its limits are, and ideally a small case study (anonymized) of where the contributor's team has used the skill in production.

A 2026 example: a strategist's team at a regional bank had built a skill for extracting structured fields from PDF mortgage applications — a problem that almost every financial-services agent has to solve, with non-trivial edge cases around handwritten signatures, scanned-and-rescanned documents, and variable field labels. They contributed the skill to the repo: the prompt, the MCP-compatible OCR tool definition, a 75-case eval set built from anonymized real applications, and a README that named the failure modes the team had hit in production. The contribution was reviewed, merged, and within four months had been used by twelve other financial-services teams (per the GitHub stars and forks). The strategist's team became the go-to reference for this problem domain in the agent community.

How to start contributing

The strategist does not need to contribute the headline skill on day one. The path that works: start by filing high-quality issues on existing skills (real bugs, eval cases that fail, edge cases the existing skill missed); graduate to small pull requests that improve existing skills (an additional eval case, a clearer README, a bug fix); then propose a new skill once the maintainers know the strategist's quality bar.

The cadence is light. A serious contributor spends two to four hours a month on the repo. The visibility is high: maintainers and peers see the strategist's commit history, and the strategist's GitHub profile becomes a portable evidence file the strategist carries to every future job, customer conversation, or conference selection.

What the company gets

The Agent Skills repo produces durable assets too. Inbound recruiting from engineers who saw the strategist's commits and want to work on a team that contributes to the field. Customer trust signals when the company's lead architect's name appears on a public agent skill the customer's team uses. Industry credibility when a new agent standard is being discussed and the company's strategist is one of the names at the table.

The company should also think about whether its own internal skills should be contributed back. The decision is product-strategic: a skill that is part of the company's differentiated moat probably stays private; a skill that is undifferentiated infrastructure (PDF extraction, ticket triage, calendar reasoning) probably gets more value from being public than from being kept private. The legal review is real but not heavy — most companies have an OSS contribution policy by 2026, and the strategist should learn it before submitting.

Venue Three: Langfuse OSS Contribution

Langfuse is the open-source agent observability platform that, by mid-2026, has emerged as the de facto OSS standard for tracing, eval, and prompt management in agent programs. It complements OpenTelemetry on the lower layer and is the layer most teams' weekly trace review actually happens on. The project's GitHub repo (github.com/langfuse/langfuse) accepts contributions, has an active community, and visibly elevates contributors into the project's documentation, blog posts, and conference talks.

Why Langfuse and not LangSmith or Helicone

LangSmith (LangChain's hosted platform) is closed-source and does not accept community contributions in the same way. Helicone has an OSS variant but a smaller community gravity. Langfuse has the OSS license, the contributor community, and the maintainer responsiveness that make it the highest-leverage place for a strategist to contribute. The strategist whose company uses any tracing platform should consider whether the contributions they would make to Langfuse — even if they primarily use Langfuse internally — would compound into the same external visibility benefits.

What a contribution looks like

Three contribution patterns for the agent strategist:

• Integration improvements. The Langfuse SDK supports many agent frameworks but coverage is uneven. A strategist whose team uses Mastra or a less-supported framework can contribute the missing instrumentation, with tests and documentation. This is the easiest entry path.
• Eval methodology contributions. Langfuse's eval module is extensible. A strategist whose team has developed a specific eval pattern (LLM-as-judge with a calibration set, multi-turn coherence scoring, tool-call-accuracy measurement) can contribute the pattern as a reusable evaluator. This is higher visibility because the contribution is conceptual and the maintainers often write about it.
• Documentation and case studies. Langfuse maintains a "how we use Langfuse at X" series of case studies. A strategist whose team has a non-trivial deployment pattern can contribute a case study. This is the lowest technical bar and the highest narrative reward.

The strategist who picks one of the three and ships a meaningful contribution within ninety days has built a public artifact that compounds. The strategist who plans to "contribute someday" never does. The discipline is to pick one and ship.

What the company gets

Langfuse contributions produce a specific kind of credibility: operational depth. A strategist whose name appears on Langfuse pull requests is signaling "this person runs a real agent program; they know what an eval is; they have hit production edge cases; they have opinions other practitioners want to read." Inbound recruiting from operations-oriented engineers spikes. Customer trust on the operations layer is automatic.

The company also gets a feedback loop. Contributing to the tools the team uses means the team's needs are heard by the maintainers, the team's edge cases inform the roadmap, and the team's lock-in to a moving open-source project becomes less risky because the company is part of the project's direction.

Cadence and Portfolio

The strategist who tries to do all three venues at full intensity in the first quarter will burn out and ship none. The discipline is a sustainable portfolio over a year, with one venue as the primary and the others as secondary.

A working twelve-month portfolio

1. Month one to three. Pick the primary venue. Most strategists in mid-2026 should pick AI Engineer Summit because the talk is the highest-leverage single artifact. Write the blog post that becomes the talk. Submit the talk proposal. While waiting on the selection committee, file three issues on Agent Skills and three issues on Langfuse to build the contributor history.
2. Month four to six. If the talk is accepted, the next three months are talk preparation. Write the slides. Practice the talk with the internal team. Submit a pull request to Langfuse (the integration improvement is easiest). Continue filing Agent Skills issues.
3. Month seven to nine. Deliver the talk. The week after the talk, publish the full blog post that elaborates the talk content. Submit the first new Agent Skill contribution based on a problem the team has solved.
4. Month ten to twelve. Harvest the inbound. Hiring funnel introductions from people who saw the talk. Customer references built on the public artifacts. Sponsor a new strategist on the team to give the next talk; pass the venue down so it does not stay personal to the strategist.

Total time investment: roughly four to six hours a week averaged over the year, with peaks of fifteen hours a week in the month before the talk. This is real time. It is not free. The strategist who does the work treats it as part of the job; the strategist who treats it as a side hustle does it badly. The Marketing team's offered help (slide template, blog editing, conference logistics) is genuine leverage; use it.

What success looks like at twelve months

• One conference talk with a YouTube recording and slide deck available publicly.
• One substantial open-source contribution to Langfuse with the strategist as named contributor.
• One published agent skill in the Agent Skills repo, used by other teams.
• Two or three technical blog posts on the company blog or a personal Substack.
• A measurable lift in inbound senior-candidate quality (the hiring team can attest).
• One or two named customer references that cite the public work as a trust signal.

This is the strategist's external evidence file. It is durable. It is searchable. It travels with the strategist to the next job and stays as a recruiting and customer-trust asset for the current company.

Three Anti-Patterns to Avoid

Anti-pattern one: the LinkedIn influencer pivot

The strategist who starts posting daily on LinkedIn with rage-bait takes ("AGENTS ARE NOT JUST CHATBOTS — here's why your company is about to lose to the competition") gains short-term engagement and long-term reputation damage. The community can tell the difference between substance and performance. The strategist who posts a thoughtful, specific technical observation once or twice a week compounds; the strategist who posts daily contrived takes accumulates noise.

Anti-pattern two: the off-message moonlight

The strategist who builds an external brand that contradicts or competes with the company's positioning creates an unmanageable problem. If the company sells a closed-source agent platform and the strategist publishes a series on "why open-source is the only ethical approach to agent infrastructure," the company is reading the strategist's posts as a resignation letter that has not been mailed yet. The alignment conversation in the Marketing meeting is not optional.

Anti-pattern three: the vanity-metric chase

The strategist who optimizes for follower count, conference invitations, or podcast appearances will drift toward content that produces those metrics — generic, fast, repetitive. The strategist who optimizes for one durable artifact per quarter (a real talk, a real PR, a real skill, a real case study) builds an evidence file that outlasts the metrics.

Story: The Strategist Who Spoke at the Summit

A strategist at a mid-cap healthcare-tech company spent three quarters of 2025 building an agent program. The program triaged provider-network claims, ran on Claude and a custom MCP server, and shipped to production in October. By December the eval set was at 1,200 cases, the trace review meeting was a weekly ritual, and the agent had handled 340,000 claims with a 97.4% accuracy rate and a documented audit trail that satisfied the company's PHI obligations.

In January 2026 the strategist applied to speak at AI Engineer Summit. The talk was titled "Shipping an Agent Into a HIPAA-Regulated Workflow: the Eval Set That Got Us Through Compliance Review." The submission was accepted in February. The talk was given in June.

The day after the talk, the strategist's company received fourteen inbound recruiter messages from senior engineers asking if the company was hiring. The strategist's LinkedIn DMs filled with peer strategists at other healthcare-tech companies wanting to compare notes. Three customer prospects who had been stalled in the procurement cycle accelerated. The talk's YouTube video, by the end of 2026, had 47,000 views.

The strategist's CEO sent a note: "Best $4K in conference sponsorship we've ever spent. Please find another conference to speak at next quarter." The strategist had not changed the program. The strategist had translated it into a story that an audience could hear, and the audience had responded.

The strategist's external voice does not create the substance; it surfaces the substance that already exists. The program has to be real first. Once it is, the public voice is the leverage that compounds the program's value into hiring, customer trust, and industry positioning. The strategist who waits to build before they speak gets exactly one shot to do it right. Most strategists ship that shot well.

This Is the End of the Program

This is the final lesson of the AI Agent Builders and Citizen Developers program. Five levels, fifty-some lessons, hundreds of pages of working patterns, dozens of named tools, and a single underlying claim: that the work of shipping agents into production is real, disciplined, learnable, and worth doing well. Level 1 taught the loop. Level 2 taught the eval. Level 3 taught the production deployment. Level 4 taught the platform. Level 5 taught the strategist's craft — program design, cost discipline, regulatory readiness, talent strategy, board reporting, the 2027 roadmap, and now the external voice.

The strategist who has completed the program is now a citizen of the agent-building community. The work continues, but the foundation is set. The next program in your career is the one you build inside your own company, on the patterns this program taught and the patterns you will invent because the field is still being invented.

Three closing observations from your authors:

First: the field moves fast, but the principles do not. The loop, the eval, the trace, the cost-per-run, the eval-driven promotion, the regulatory posture, the named owner, the honest report — these are durable. The specific tools will turn over (Claude 4 becomes Claude 5, Temporal gets a competitor, A2A becomes A2A-plus-something). The discipline is what makes the strategist effective regardless.

Second: the agent industry needs more strategists. There are more engineers than there are people who can translate engineering into business and back. The strategist's role is high-leverage precisely because it is rare. Becoming visible in the community is not optional self-promotion; it is contribution to a community that needs the role to exist and to be done well.

Third: the year the strategist spent on this program was not wasted regardless of the specific job outcome. The skills are portable. The patterns are reusable. The relationships built in the community travel with the strategist. The work of becoming a strategist is, in a real sense, the work of becoming the kind of professional whom companies want and whom the industry needs. That work continues forever.

Go build. Then go talk about what you built. Then go help someone else build the thing they have been afraid to start. That is what the community is for.

Key Takeaways

• The strategist's public voice is hiring leverage, customer trust, and a seat at the table where industry conventions are written. It is not a brand exercise.
• Align with the company's communications goals before publishing. Three questions for Marketing: what story is the company telling, what topics are off-limits, who reviews before publication.
• AI Engineer Summit (twice yearly, swyx and Latent Space): the highest-leverage single talk venue. 20-30 minutes on a specific production problem with real numbers. Write the blog post first; submit the talk based on it.
• Agent Skills GitHub repo (anthropics/skills): contribute self-contained skills with eval sets, MCP-compatible tool definitions, and READMEs. Start with high-quality issues, graduate to PRs, then propose new skills.
• Langfuse OSS (langfuse/langfuse): the highest-leverage OSS observability project. Contribution patterns: integration improvements, eval methodology, case studies. The contribution signals operational depth.
• Cadence: four to six hours a week averaged over a year. One conference talk, one substantial OSS contribution, one published agent skill, two or three blog posts. Total time is real but sustainable.
• Anti-patterns: the LinkedIn influencer pivot, the off-message moonlight (contradicting company positioning), the vanity-metric chase. The strategist optimizes for durable artifacts, not engagement.
• What success looks like at twelve months: a YouTube talk recording, a named OSS contribution, a published agent skill, a measurable lift in senior-candidate inbound, one or two customer references citing the public work.
• The program's foundation: the loop, the eval, the trace, the cost-per-run, the eval-driven promotion, the regulatory posture, the named owner, the honest report. Durable across tool changes.
• The agent industry needs more strategists. Visibility in the community is contribution to a community that needs the role to exist and to be done well. Go build, then go talk about it, then go help someone else build the thing they have been afraid to start.