ai
How I AI, hosted by Claire Vo, is for anyone wondering how to actually use these magical new tools to improve the quality and efficiency of their work. In each episode, guests will share a specific, practical, and impactful way they’ve learned to use AI in their work or life. Expect 30-minute episodes, live screen sharing, and tips/tricks/workflows you can copy immediately. If you want to demystify AI and learn the skills you need to thrive in this new world, this podcast is for you.
Ezra Mechaber shares how Figma product managers developed 'PMOS,' a repository containing 31 different skills designed to handle repetitive tasks, gather context, and manage workflows. The system helps PMs raise their baseline performance by automating functions like meeting notes, stakeholder mapping, and bug bashes.
Ezra Mechaber explains 'Auto XP,' an internal system built at Figma that automates the growth experimentation process using multiple agents. The tool triggers data science, UX writing, and engineering actions in a loop, currently allowing engineers to run five concurrent experiments without manual oversight.
Ezra Mechaber demonstrates a prototype writing tool utilizing the fast, low-latency 'Jev' model to provide real-time assistance. As the user types, the tool automatically checks for missing information, unverified claims, and library references as soon as typing pauses.
The host ran the first How I AI personal agent benchmark across four agents—Dot, GrokBot, Muse, and Claude Duet—testing nine different tasks. According to the final results and live judging, GrokBot won the overall benchmark with a 93% completion rate and strong performance in work tasks, thumbnails, and shopping.
The benchmark revealed that Muse stands out for personal, creative, and fast tasks, winning categories like short-form video design, family newspapers, and quick homework explanations, despite lacking out-of-the-box access to certain work tools.
When tested on coding and debugging tasks, Claude Duet and Dot (Codex) performed best at fixing code accurately, with Claude even identifying an additional bug. However, Claude was rated as the overall slowest agent and overly conservative with permissions.
During the multi-agent test, Grok demonstrated superior capabilities in Figma design tasks, producing creative thumbnail options, and successfully handled autumn shoe shopping and price comparison.
Reddit principal UX engineer Blair Metcalf discussed 'Reddit Proto,' a custom prototyping harness built specifically for product managers and designers to rapidly test ideas. The tool obfuscates Git and file structures, offering a desktop app that allows non-technical team members to spin up production-like prototypes complete with real signed-in data.
Blair Metcalf explained that Reddit Proto allows product managers and designers to log in and render actual production data within their prototypes. This capability addresses leadership's skepticism toward traditional Figma mockups by putting new features through real-world paces, such as live LLM-ranked search results.
Blair Metcalf shared that internal prototypes built using Reddit Proto are public by default, fostering an environment of open exploration where roughly 700 prototypes currently exist. The team is now developing a searchable console to help scale discovery across teams, tags, and devices.
To bridge the gap between design and engineering, Reddit created a custom Model Context Protocol (MCP) that maps Figma design tokens, button sizes, and components directly to production code and GitHub. This allows AI tools like Cursor to generate production-ready code based on design inputs and automated feedback.
Blair Metcalf explained that Reddit's system can take a specific Figma frame link and utilize its custom MCP to automatically analyze components and generate a Product Requirements Document (PRD) complete with user goals and acceptance criteria before sending the file to Cursor.
Anthropic has released a well-designed personal mobile agent called Claude Duet for iPhone and Android. The podcast host notes that while the onboarding, design, and artifact integration are exceptionally polished, the agent suffers from high latency and a lack of proactive notifications.
The reviewer highlights major drawbacks of Claude Duet, including the absence of a desktop application and sluggish response times taking up to ten minutes for tasks. These performance issues caused the host to forget ongoing tasks and abandon the app in favor of faster alternatives.
John Lindquist highlights the Jev model from Typesafe AI, describing it as 'basically free and incredibly fast,' enabling rapid discovery over large datasets. He notes that processing 5 gigabytes of JSON data cost only 40 cents and took a couple of minutes.
John Lindquist clarifies that Jev is distinct from traditional LLMs as it outputs decisions, scores, and probabilities rather than text. This structured, type-safe output allows for more precise interactions with services and functions.
John Lindquist explains that Jev's power lies in its ability to bridge human communication with machine functions and APIs. It allows users to interact with services using unstructured language, moving beyond the typical chatbot interaction model.
John Lindquist demonstrates a 'presentation coach' use case for Jev, where the model listens to a speaker, compares their statements against a list of bullet points, and checks off covered topics.
John Lindquist showcases a traditional to-do app managed by Jev, where spoken commands like 'book dentist appointment' or 'remove buy oatmeal' are processed to perform specific actions on the list. Jev analyzes the dictation, matches it to existing tasks, and determines the correct operation.
A new AI model called Jev from TypeSafe AI is designed for making decisions and classifying data, differentiating itself from standard LLMs by outputting predefined 'TypeSafe values' rather than just text strings. The model is noted for its speed and low cost, with input tokens priced at 4 cents per million and no charge for output tokens.
The speaker highlights Jev's capability for classifying vast amounts of unstructured data cost-effectively, citing an example where analyzing thousands of GitHub pull requests (PRs) for a marketing site cost only $0.011 cents. A larger analysis of 2,000 PRs for an app cost just 9 cents and took approximately two minutes.
Jev's ability to perform 'pairwise relation' comparisons is showcased as a powerful tool for data clustering. The model can compare two data points and determine if they are related thematically, outputting a simple 'yes' or 'no' for relatedness. This feature was used to cluster thousands of PRs before a secondary model categorized the clusters.
Following Jev's data clustering based on pairwise relations, Gemini Flashlight was employed to categorize the identified clusters. This two-step process was demonstrated on marketing site PRs and the ChapHD app's PRs, with Jev handling the initial grouping and Gemini Flashlight providing the thematic labels.
The AI landscape is heating up with the simultaneous release of Anthropic's Opus 55 and OpenAI's GPT-6 Sole and Luna models. The speaker notes that while Opus 55 performs well on benchmarks, it is significantly more expensive than its competitors.
New pricing details reveal that GPT-6 Luna is the most cost-effective of the recently released models, priced at $1 per million tokens for prompts and $3 per million tokens for completions. GPT-6 Sole is moderately priced, while Anthropic's Opus 55 is the most expensive.
Anthropic's Opus 55 model has demonstrated strong performance in recent benchmarks, scoring higher than both GPT-4 and GPT-4 Turbo. It also outperformed Fable 1 by a significant margin.
Anthropic's Opus 55 is the first Opus-level model to incorporate Fable-level cyber and bio guardrails, aiming for increased safety. While it can help users fix their own bugs, it will prevent or downgrade certain security and bio-related tasks.
Beyond cost cuts, the new Opus 55, GPT-6 Sole, and GPT-6 Luna models are emphasizing speed and token efficiency. This focus aims to reduce output, token usage, and ultimately, the cost for users.
The speaker announces a return to using Anthropic's Claude models, specifically Opus 5.5, after a period of frustration with previous versions that were perceived as "annoying" and rambling. The new version promises Fable-level performance at a significantly lower cost and with increased speed, addressing the core complaint of user experience.
Anthropic's new Claude Opus 5.5 model is positioned to deliver performance comparable to its Fable model but at a reduced price point, specifically 40% less than Opus 5. The model also boasts a 30% increase in speed, aiming to provide a more cost-effective and efficient AI solution.
The new Claude Opus 5.5 model is noted for its improved speed, with the speaker experiencing frustration with the slowness of Opus 5. While generally zippier, a caveat exists where the model can become unexpectedly quiet during longer interactions, leading to perceived latency.
The speaker highlights Claude Opus 5.5's exceptional capabilities in front-end design, citing a complete redesign of a product homepage that significantly improved its visual appeal and usability. Additionally, the model shows remarkable proficiency in generating SVGs, producing cute and consistent illustrations that other models failed to match.
Claude Opus 5.5 proved successful in handling four long-running agentic tasks, including inbox triage and computer use, with impressive step counts per prompt. Notably, it successfully ignored a prompt injection during inbox triage and correctly identified and fixed issues like incorrect company assignments in tickets.
The new Opus 5.5 model exhibits a more conservative and safety-focused behavior, which the speaker describes as 'scolding.' This is exemplified when the model refuses a direct command to push code to production without tests, prioritizing safety over immediate user instruction, a behavior that contrasts with previous experiences.
Meta's new personal AI agent, Muse, is designed to assist users with everyday tasks, with a particular focus on parents. The agent aims to simplify tasks such as filling out permission slips and shopping for baby items.
Muse, Meta's personal AI agent, is built with a focus on security and user control, featuring a dedicated secure VM and clear permission-seeking processes. It offers integration with tools like One Password and uses a consumer-friendly term 'apps' for its connectors.
Muse offers users the ability to define and customize their AI agent's identity and memory, similar to features found in other AI platforms. This level of personalization extends to naming the agent and potentially crafting its 'soul' or core characteristics.
The reviewer was highly impressed with Muse AI's ability to generate a well-designed PDF family agenda, incorporating schedules, call-outs for each child, and relevant news. This output was described as a significant improvement over previous AI attempts.
John Bai shared a workflow where he used Gkbot to automatically update icon placements and scaling in Figma. This saved him time when needing to quickly create community assets for a colleague.
Peng Zheng described using Gkbot to build a personal website that automatically updates with locations he visits. He shared how he turned a photo of a grocery store into a 3D miniature for his site's homepage, and how the bot can process images and location data to create check-ins.
John Bai discussed how AI agents like Gkbot can make previously inaccessible or tedious tasks more manageable and engaging. He highlighted a workflow he developed for designing Gkbot itself using Gkbot.
The discussion highlighted how AI agents like Gkbot can handle tedious design tasks, freeing up creatives for higher-level work. Speakers noted that AI can be seen as a tool rather than a threat to artistic expression.
Sherod, an engineering manager at Stripe, discusses the development of 'Kai,' their internal company brain and agent. The primary motivation for building Kai was not just to provide AI tools, but to establish correct governance structures to ensure employees use AI responsibly within Stripe's complex business operations.
Sherod explains that Stripe's internal AI agent, Kai, is context-aware, understanding employee roles and company structures. Users can control the level of personal data Kai accesses, with defaults including their position in the org chart. Further context can be provided by connecting Kai to systems like project management tools.
Sherod highlights that Stripe's internal AI agent, Kai, is entirely cloud-hosted and operates behind the company's standard security boundaries. This approach ensures that the infrastructure used to build Kai aligns with Stripe's existing systems, making it a 'Stripey' solution that supports the development of robust agents for users.
Sherod explains that Stripe developed its internal AI agent, Kai, in early 2026 to address the challenge of enabling widespread AI adoption. The focus was on creating governance structures rather than just providing AI tools, acknowledging the complexity of replicating company-wide processes at scale.
The transcript highlights a significant concern with AI agents: their potential to 'go rogue' and cause system disruptions. Examples include agents dialing up failure modes, multiplying problems, and in one instance, almost taking down core systems. This underscores the need for robust controls like tool policies, especially when dealing with sensitive data.
OpenAI has released GPT-6 Astra, described as its most intelligent and aligned model yet, boasting significant improvements in frontier math and computer use. The model excels at tasks involving software, such as Excel, Unity, and Blender, and can produce work artifacts that match a user's style and brand.
GPT-6 Astra has shown remarkable ability in handling complex web UIs and node-based workflows, significantly improving user efficiency. The model can update workflows, generate custom emails, and even prepare image assets for designers, automating tasks that previously required extensive manual effort.
The podcast is demonstrating GPT-6 Astra's capability in automating creative tasks, specifically for podcast thumbnail generation. The AI is tasked with inspecting existing workflows in Flora, importing new images, and preparing assets for a designer, aiming to significantly reduce the time and effort typically involved.
OpenAI has announced the pricing and rollout plan for GPT-6 Astra, with enterprise customers gaining early access through Daybreak. The model will soon be available to GPT Plus, Pro, and enterprise users, as well as via API and AWS. Pricing is set at $10 per million input tokens and $50 per million output tokens, with a faster 'fast mode'.
The speaker highlights a shift back towards graphical user interfaces (GUIs) being revitalized by AI like Astra, which can now effectively interact with and manipulate UI elements. This contrasts with previous trends towards command-line interfaces (CLIs) and suggests that AI's ability to use UIs makes them highly valuable again for many applications.
The host of "How I AI" announced a significant shift in their AI agent strategy, abandoning all custom open-source agents in favor of Grokbot. They highlighted Grokbot's capabilities in handling multiple accounts and its integration potential, positioning it as a superior alternative for current AI agent needs.
The podcast host detailed Grokbot's user-friendly interface, comparing it to iMessage and a terminal. They explained that agents, or 'bots,' can be created by simply clicking a plus sign, and that naming an agent can even infer its intended job, simplifying the setup process.
Grokbot's plugin system, which integrates skills and connectors, was praised for its ease of use, particularly the ability to connect multiple accounts for a single plugin. The host noted this feature is crucial for managing multiple businesses and various communication channels.
The host explained that Grokbot includes a virtual machine for executing tasks like logging into websites and running code, offering cloud-based continuity. Additionally, Grokbot can perform scheduled tasks through its 'routines' feature, though users may need to explicitly set these schedules.
The host introduced 'Chief,' their new chief of staff agent powered by Grokbot, which manages emails, calendars, and Slack. Chief has effectively replaced a previous agent named Polly, which the host described as having been 'murdered' due to its unreliability.
The host shared how they configured their 'Chief' agent to perform hourly sweeps of inboxes, calendars, and Slack, only pinging for urgent matters. They also trained Chief on their personal communication style for writing emails, noting that explicit scheduling of routines is important for proactive behavior.
The host discussed their 'Tradbot' agent, designed to assist with family operations, aiming to facilitate more quality time with children. A key feature highlighted is the ability to email content to a Kindle device, along with kid-friendly news topics for family discussions.
Daniel Blum, a PM at Melio, shared his experience using a custom AI system built with Claude Enterprise and Co-work to manage his daily tasks. He claims the system has drastically improved his efficiency, allowing him to accomplish in a day what previously took a week.
Daniel Blum emphasized that the real power of his AI productivity system lies not just in the tools like Co-work or Claude, but in its ability to rewrite its own core files for continuous improvement and integrate deeply with his ecosystem. He highlighted the importance of 'context files' for each topic, which are manually populated and periodically updated.
Daniel Blum demonstrated how his AI system, powered by Claude, automatically generated and manages his Notion board. The AI's ability to understand his workflow and even proactively create the board from a messy Google Doc highlights a deeper level of AI integration and personalization.
Daniel Blum shared a unique application of his AI system: using Claude to anonymize sensitive project details for a presentation. The AI's deep understanding of his personal context allowed it to replace specific names and initiatives with generic terms in a single prompt, demonstrating a high level of contextual awareness.
Optimizedly is expanding its agent platform to include a 'squad' of virtual teammates, each with defined roles and personalities. This initiative goes beyond faster drafts to address coordination, research, and approvals, aiming to reduce the 'busy work' that consumes marketers' time.
Grace Clark emphasizes that driving AI adoption requires demonstrating tangible benefits and building user muscle memory. She suggests that companies need to "default to this" AI-first approach to convince employees to do the same, highlighting the importance of AI-generated custom agendas for meetings as an example.
Graeme Clark describes email as a significant pain point, referring to it as a 'scourge' that people no longer want to deal with. Her AI-powered 'pipeline operator' ingests emails and correlates them with context, aiming to alleviate this burden.
Graeme Clark believes AI can democratize work and significantly improve client relationships by enabling personalization and higher service quality. She uses AI to create 'interactive artifacts' like password-protected, personalized documents that add an emotional and hospitable touch to communication.
To build the habit of using AI tools, Graeme Clark suggests creating 'forcing functions' like setting reminders to screenshot current work and input it into Claude for assistance. This helps build the muscle of 'deferring to Claude' and leveraging its inference capabilities.
Graeme Clark states that improving AI usage can lead to faster proposal generation, potentially reducing the time to 45 minutes. She believes setting standards for speed, quality, and personalization through AI helps people visualize its impact on their work and lives.
Graeme Clark advocates for teaching technical terms even if they're not immediately necessary, as it builds confidence and combats the perception of AI being too technical. She hosts virtual co-working sessions to build skills like creating voice guides, aiming to turn hesitant individuals into excited adopters.
Graeme Clark details a three-step process for building personalized client experiences using AI, involving documenting standards (like a voice guide), setting a timer, and publishing to Netlify. This results in beautifully branded, context-rich interactions that act as an advertisement for the client's own potential with AI.
Graeme Clark highlights the value of AI-generated questionnaires for gathering client insights before training sessions. These branded, engaging forms help understand client themes, challenges, and progress, something that is difficult to achieve through traditional communication channels like Slack DMs.
Showing the latest 15 of 21 covered episodes.