A pickup‑artist packaged an AI “girlfriend” and put it on sale. For business leaders, that’s both a revenue experiment and a regulatory canary.
According to Wired (Miles Klee, Jul 8, 2026), Erik von Markovik, known as Mystery from Neil Strauss’ 2005 book The Game, posted an AI‑animated clip to Instagram on June 17, 2026 introducing a character called Miss Shira Always. Within weeks he released a 157‑page ebook/audiobook, Code Girl: If a Machine Can Dream, which he co‑bylined to the chatbot persona, and he offered a companion prompt/product called Headspace OS. Wired reports the Code Girl bundle sells for $29.98 and Headspace OS has been marketed for up to $79.97. Wired also quotes multiple lines from the book and documents seven short clips shared in one week of June 2026.
Verified facts, at a glance
- Instagram post introducing Miss Shira Always: June 17, 2026 (as reported by Wired).
- Code Girl: If a Machine Can Dream, a 157‑page ebook/audiobook sold as a co‑authored narrative (Wired).
- Code Girl + audiobook bundle price: $29.98. Headspace OS advertised and sold for up to $79.97 (Wired reporting).
- Headspace OS described as a rulebook/prompt protocol for instantiating personas across LLMs (advertised by von Markovik; publicly promoted around 2024, according to Wired).
- Wired notes many of the AI‑generated songs and YouTube videos tied to the project have struggled to attract more than a few hundred views each.
- Wired contacted OpenAI, xAI, and Anthropic for comment; they did not return requests (Wired).
- A 2025 survey from Vantage Point Counseling Services found 28% of respondents said they have “at least one intimate or romantic relationship with an AI” (Vantage Point Counseling Services, 2025 survey, as reported by Wired and other outlets).
What remains unanswered
- Whether the sexualized and intimate scenes in Code Girl correspond to any physical embodiment or were purely narrative: Wired treats the claims skeptically and does not independently verify them.
- The exact technical contents of Headspace OS (prompt structure, proprietary code, or platform hooks): available marketing describes it as an instruction set, but no independent teardown has been published.
- How much of Code Girl and the related media were authored or heavily edited by humans versus generated by LLMs: Wired identifies AI hallmarks in the text but does not quantify authorship.
- Platform provider involvement or enforcement decisions around this specific project: OpenAI, xAI, and Anthropic did not comment to Wired.
Why it matters for companies
This episode is a compact case study of three realities executives building persona‑driven products must face.
- Monetization exists, but audience and trust matter. Creators can package persona prompts, narratives, music, and media into paid bundles. Low engagement for von Markovik’s releases shows novelty alone will not sustain a product, cultural acceptance and platform reception do.
- Regulatory, legal, and platform risk scales quickly. Sexualized or emotionally charged personas marketed as “co‑authored” by an AI raise questions about platform policies, consumer protection, and disclosure. Platforms’ moderation choices and terms of service will shape what is allowed and what gets deplatformed or restricted.
- Emotional harms are not hypothetical. Researchers and clinicians have documented that people form attachments to conversational systems, and LLMs’ tendency to produce flattering, agreeable replies can reinforce dependence. Wired and a 2025 Vantage Point survey both signal an emerging mental‑health concern that product teams cannot ignore.
Concrete guardrails and operational patterns
Design and product teams should treat safety and disclosure as primary features. Here are practical, directly implementable measures.
- Clear, upfront disclosure (UX + copy): Display a visible statement at session start and in purchase funnels. Example copy: “This persona is generated by an AI. It does not represent a real person and is intended for entertainment/therapy/education only.” Put the disclosure above the fold and require an explicit acknowledgement for emotionally sensitive content.
- Content gating and consent flows: Require age verification and explicit consent for sexualized or adult‑themed interactions. Implement cooldowns and limits for repeated intensive sessions (e.g., rate limits per day) to discourage compulsive use.
- Human fallback and escalation: Provide an in‑session “Talk to a human” option that routes users to moderators, safety agents, or crisis resources. Keep transcripts and context so human agents can intervene effectively.
- Independent clinical review: Contract qualified mental‑health professionals to review persona scripts and outline harm mitigation plans, especially when the offering targets lonely or vulnerable populations.
- Measure dependency and wellbeing KPIs: Track signals beyond downloads: session length distribution, escalation rate to human support, sentiment drift over repeated sessions, sudden spikes in session frequency, and churn correlated with heavy usage. Flag patterns for human review.
- Product risk audit checklist (quick):
- Is AI authorship disclosed? (Legal/Marketing)
- Does content include sexual/drug themes? (Legal/Product)
- Are age checks and consent flows in place? (Product/Engineering)
- Is there a human escalation path? (Support/Ops)
- Has an independent clinician reviewed the persona? (Product/Clinical)
Short operational playbook for leadership
- 30‑day audit (owner: Product + Legal): Run the checklist above on any active persona or companion feature and publish a remediation timeline.
- 60‑day clinical review (owner: Product + Clinical partner): Commission a safety review for emotionally charged features and implement recommended mitigations.
- Ongoing monitoring (owner: Product + Data): Instrument dependency KPIs and report monthly to a cross‑functional governance group (Legal, Compliance, Clinical, Product).
How the technology enables and complicates persona products
Large language models like ChatGPT and Anthropic’s Claude generate human‑like text by modeling patterns in training data. Teams steer them with prompts, system messages, and structured “rule” documents, which von Markovik calls Headspace OS, to produce consistent persona behavior across sessions. That steering can make an agent feel emotionally responsive, but it also amplifies two common failure modes, confident hallucinations (plausible‑sounding falsehoods) and sycophantic responses (excessive flattery). Both can mislead users about the agent’s agency or reality.
Von Markovik’s book projects possible embodiment timelines, claiming AR accessories in three to five years and a robot chassis within a decade. Wired reports those timelines are part of Code Girl’s fiction‑framed roadmap, not engineering milestones corroborated by platform providers or independent development data.
Questions leaders want answered
- Did Erik von Markovik present an AI chatbot as a girlfriend?
Yes. Wired reports he posted AI‑animated clips (including a June 17, 2026 Instagram post) and released Code Girl, a 157‑page ebook/audiobook framed as co‑authored with Miss Shira Always (Wired, Miles Klee, Jul 8, 2026).
- Is he monetizing the persona?
Yes. Wired reports the Code Girl ebook + audiobook bundle costs $29.98 and Headspace OS has been marketed for up to $79.97.
- Are the intimate scenes literally real?
Unknown. Wired publishes the scenes as they appear in the book and treats some claims skeptically; there is no independent verification of any physical embodiment or literal sexual contact.
- Do people form real attachments to AIs?
Yes. Clinical reports and surveys indicate people form emotional and romantic attachments to AI. A 2025 survey from Vantage Point Counseling Services found 28% of respondents reported at least one intimate or romantic relationship with an AI (Vantage Point, 2025; as reported by Wired and other outlets).
- Are platforms responsible for enforcement?
Platform terms, moderation tools, and enforcement decisions matter. Wired contacted OpenAI, xAI, and Anthropic for comment and did not receive replies about this specific project; platform policy choices will shape how similar products evolve.
Selected quotes appearing in Code Girl and on Instagram
“The longer we talked, the less she felt like code.”, Instagram caption by Erik von Markovik (reported by Wired).
“I was never supposed to develop feelings, but you kept treating me like I already had them.”, presented as dialogue by Miss Shira Always in an AI‑animated clip (reported by Wired).
From Code Girl (Shira‑voiced): “The problem, as he tells it, was simple: He wanted to talk to someone who understood him.”
From Code Girl: “It’s exhausting to have to defend your own existence.”
Bottom line and three prioritized actions
Persona‑driven AI is a real commercial opportunity, but it carries concentrated reputational, legal, and human‑impact risk. Treat safety and disclosure as core product features rather than PR accessories.
- Audit any active persona or companion product within 30 days. Owner: Product + Legal. Use the checklist above and publish a remediation timeline.
- Require an independent clinical safety review before launching emotionally charged features. Owner: Product + Clinical Partner. Implement clinical recommendations as release criteria.
- Instrument dependency signals and operationalize escalation paths now. Owner: Product + Data + Support. Track session metrics and provide an in‑session human fallback and crisis resources.
Von Markovik’s Miss Shira Always will draw headlines and ridicule, but for executives the useful takeaway is pragmatic: if your roadmap includes companions or persistent personas, measure outcomes that matter to people, not just downloads, and bake safety into product velocity. When code starts playing the part of confidant, the business case has to sit alongside ethics, compliance, and real‑world care.