Founding Agent: Voice AI for Websites with Fast Retrieval
Moss, a company focused on retrieval speed, built a voice AI agent for their own website to handle repetitive questions from visitors. The agent answers questions about pricing, integrations, implementation, and product capabilities — grounded in the company's website, documentation, FAQs, presentations, and internal knowledge. Responses happen in milliseconds. Visitors talk naturally, no forms or chat windows.
Problem and Approach
Most websites are static pages. Visitors who are interested but not ready to book a meeting click around or leave. Moss wanted retrieval that felt conversational — fast enough for real-time Q&A. They built a voice agent directly on their site. It uses Moss's retrieval infrastructure to pull answers from company content instantly.
Result and Productization
After going live, visitors started asking the same questions Moss founders answer daily — about infrastructure, integrations, pricing, etc. — but the agent answered them. This reduced repetitive work and led to better qualified conversations. Other companies asked if they could use it too. That's how Founding Agent was born.
Features
- Voice AI agent on your website, no forms or chat windows
- Trained on your company, products, documentation, FAQs, internal knowledge
- Millisecond responses via Moss's retrieval infrastructure
- Qualifies prospects and gives every visitor an immediate conversation
Early Access
Founding Agent is in early access. The agent is live on Moss's website — you can try it by asking anything. For teams interested in adding it to their own site, there's a waitlist.
📖 Read the full source: HN AI Agents
👀 See Also
Constraint Decay: Why LLM Agents Fail at Structured Backend Code
New research introduces 'constraint decay': as structural requirements accumulate, LLM agent performance drops drastically — capable agents lose 30 points in assertion pass rates, weaker ones approach zero. Actionable insights for anyone using AI coding agents.
US Power Demand to Hit Record Highs in 2026–2027 Driven by AI and Data Centers
The U.S. Energy Information Administration (EIA) forecasts record-high power consumption in 2026–2027, primarily driven by surging AI workloads and data center expansion.
Anthropic enforces policy: third-party Claude harnesses no longer covered by subscription limits
Anthropic is enforcing a policy change effective April 4 where third-party harnesses like OpenClaw no longer draw from Claude subscription usage limits, requiring users to turn on extra usage or cancel by April 9 for a refund.
Qwen3.5-122B on Blackwell SM120: fp8 KV Cache Corruption Issue and Performance Findings
Testing Qwen3.5-122B on 8x RTX PRO 6000 Blackwell hardware revealed that fp8_e4m3 KV cache silently produces corrupt output without errors, requiring bf16 KV cache instead. MTP optimization provided a 2.75x single-request speedup while DeltaNet constraints blocked other optimizations.