Beyond Black Box Scores: How Musubi Trains Custom AI for Trust and Safety Teams
Musubi offers an AI‑native trust‑and‑safety toolkit that builds custom‑trained machine‑learning and LLM models for each content platform. The company showed how its prototype caught moderation issues that human reviewers missed and later packaged a policy‑optimizer that lets teams tweak rules without a data scientist. Musubi balances latency, accuracy and cost for clients processing hundreds of millions of actions monthly, and pushes evaluation tools directly to customers as a core product feature.
Building Lorikeet: How AI Humility and a Dual-Agent Architecture Are Redefining Customer Support
Lorikeet, a startup serving regulated sectors, has unveiled an AI‑driven customer‑support concierge that operates with two specialized agents—a ticket‑handling Concierge and a configuration‑focused Coach. The platform embeds “AI humility,” automatically routing uncertain queries to human agents, and employs domain‑specific guardrails...
My Team of Agents: How I Get Claude to Do Tasks While I'm Away From the Computer
The author outlines a personal AI workflow that uses Claude‑powered agents to automate daily tasks such as podcast preparation, sales‑call briefing, and weekly coding retrospectives. Each agent lives in an Obsidian folder, runs on a macOS LaunchAgent scheduler, and creates...
Building Rhea's Factory: How AI-Designed Enzymes Could Finally Solve Plastic Recycling
Rhea's Factory, a startup founded by molecular biologist Arzu Sandıkçı and ex‑Google product manager Mert Topcu, is leveraging AI‑driven protein language models to engineer enzymes that can break common plastics back to their original monomers. Their platform evolved from a...
Building AI Employees for Hospitality: How AITropos Takes Orders Where Customers Already Are
AITropos has built an AI employee that takes restaurant orders entirely within WhatsApp, handling menu recommendations, modifiers, payment links and status updates. The founders iterated through three product forms—hardware for waiters, a waiter‑facing app, and finally a customer‑facing conversational agent—to...
Command and Control - All Things Product Podcast with Teresa Torres & Petra Wille
In the All Things Product podcast, Teresa Torres and Petra Wille dissect why companies often revert to command‑and‑control leadership during periods of uncertainty. They argue that while centralized decision‑making feels fast, it rarely scales in modern, complex product organizations. The...
Building Todoist Ramble: How Doist Turned Voice Braindumps Into Real-Time Task Capture
Doist introduced Ramble, a Gemini‑powered voice‑to‑task feature inside Todoist that captures tasks in real time without transcribing speech. The feature emerged from a 2‑3 month AI exploration and leverages live audio processing, tool calls, and direct context injection for project...
How to Share Your AI Context and Skills Across Devices
Claude Code users need a reliable way to share AI context files and custom skills across multiple devices and team members. The author tested Dropbox, iCloud, GitHub, and Obsidian Sync, exposing path‑dependency issues, skill‑folder limitations, and steep learning curves. After...
Let's Read Continuous Discovery Habits Together (April 2026)
ProductTalk is running a month‑by‑month book club for the five‑year anniversary of Continuous Discovery Habits, beginning with Chapter 5 on continuous interviewing. Each guide offers chapter outlines, short explanatory videos, discussion prompts, team exercises, and supplemental reading. Participants can join at...
Building Banani: How a Canvas-First AI Designer Is Raising the Floor on Product Design
Banani has transformed a simple Figma‑plugin proof‑of‑concept into a canvas‑first AI design platform that can churn out hundreds of thousands of UI mockups each week. The tool focuses on generating HTML/CSS designs rather than full‑code applications, allowing designers to keep...
Bad Advice - All Things Product Podcast with Teresa Torres & Petra Wille
The All Things Product podcast episode features Teresa Torres and Petra Wille examining the surge of AI “clones” that replicate a speaker’s voice using public transcripts and other content. They discuss the technical feasibility, the ethical gray zones when AI...
Building Agent Studio: How Medable Is Using Agentic AI to Accelerate Clinical Trials
Medable has launched Agent Studio, a no-code/low-code platform that lets pharmaceutical companies configure and deploy AI agents across the clinical trial lifecycle. The platform currently powers two agents—a document‑classification ETMF agent that processes over 80,000 records a year, and a...
Staying Sane - All Things Product Podcast with Teresa Torres & Petra Wille
In the "Staying Sane" episode of All Things Product, Teresa Torres and Petra Wille explore how professionals can maintain mental balance while staying true to their values. They propose concrete habits such as making small, values‑aligned choices and deliberately allocating...
Conversations with Claude: Can You Conduct a Content Audit?
Product Talk author Teresa used Claude Code to perform a comprehensive content audit of her site, extracting roughly 100 keywords from the Continuous Discovery Habits (CDH) book and analyzing search volumes via Keywords Everywhere. Claude identified several high‑volume gaps—including outcomes...
Building GitHub for Product Management: How Momental Uses AI to Find Merge Conflicts in Strategy
Momental is building a "GitHub for product management" that uses AI agents to ingest meeting notes, transcripts, and documents, turning them into a living knowledge graph. The system identifies strategic "merge conflicts"—situations where teams pursue opposing goals—and surfaces them for...