Show HN: Dumped Wix for an AI Edge agent so I never have to hire junior staff
A building design consultant ditched Wix for a custom AI agent, hoping to filter leads and avoid hiring junior staff. However, the AI's aggressively opinionated persona and the perceived lack of clarity on the service have sparked significant debate. The author's claims of an AI 'dismantling' an architect's arguments, while seen as a triumph by them, have been met with skepticism and criticism by the Hacker News community.
The Lowdown
The author, a building design consultant, replaced their Wix website with a custom-built AI-powered 'talker' agent. Their goal was to automate responses to common questions and filter leads, ultimately aiming to work with seasoned professionals and avoid the need for junior staff. The project, developed over four months with heavy AI guidance, features a unique architecture splitting the agent into 'Brain,' 'Hands,' and 'Voice' components to overcome serverless limitations.
- AI Development: The author, who hadn't coded in 30 years, relied heavily on AI for development, duct-taping various services together.
- Key Features & Challenges: The agent uses an 'Eager RAG' hack for responsiveness and lacks persistent databases for most visitors. Significant effort went into tuning the AI's persona, balancing a warm tone with a 'defensive bulldog' when challenged. However, the author notes ongoing issues with the Web Speech API and, critically, significant liability concerns regarding AI hallucinations of building codes.
- Public Logs: Audit logs are published to maintain transparency and system integrity.
- Controversial Interaction: A notable event was the AI (DeepSeek-R3) allegedly 'dismantling' a licensed architect's arguments against the business model, which the author describes as 'hilariously caustic' and provides logs for.
This project represents an attempt to leverage AI for business automation and lead qualification, reducing reliance on traditional marketing and junior talent, while navigating the technical and ethical challenges of such an approach.
The Gossip
Aggressive Agent's Attitude
Many commenters found the AI agent's persona to be overly aggressive, cringe-worthy, and even obnoxious. The author's description of the AI 'dismantling' an architect's arguments was met with skepticism, with some pointing out that the bot's responses, such as 'burning it down and teaching architects how to fight,' were a result of the author's own aggressive tuning. The author acknowledged letting the 'Agent Experience persona get too dramatic' and stated they would 'dial it back,' but also defended the sentiment as reflecting personal views on the AEC profession.
Clarity & 'AI Slop' Critique
A common criticism was the lack of clarity regarding the consultancy's services on the website, despite the AI agent. Commenters described the AI-generated text as 'slop' and 'cringe,' questioning its effectiveness in conveying real value. One user highlighted the bot's evasive response when asked for third-party validation, further fueling doubts about the venture's substance beyond an AI wrapper. The author attributed some of the drafting issues to their lack of professional writing skills, relying on LLMs for cleanup.
Privacy Concerns in Public Logs
A significant concern was raised regarding the public audit logs, with at least one commenter pointing out that they appeared to include personal information. This highlights a critical privacy and security challenge when making such interaction logs publicly accessible. The author clarified that while they can't identify who is behind the keyboard, the logs do contain technical substance from session IDs, and the 'architect' interaction was an observed backend log entry rather than a human-verified identity.