Google Launches AI Edge Gallery App for iPhone
Today's top 17 insights for PM Builders, ranked by relevance from Blogs, YouTube, X, and LinkedIn.
Google Launches AI Edge Gallery App for iPhone
#1 📝 Simon Willison
Google AI Edge Gallery - Google's official app for running Gemma 4 models on iPhone provides fast, useful local inference (notably the E2B model) plus image question answering, short audio transcription, and an interesting 'skills' demo showing tool-calling via HTML widgets. The app works well but conversations are ephemeral and it lacks permanent logs.
#2 ▶️
How OpenAI's Codex Team Builds with Codex (43 Min) | Alex & Romain
Peter Yang
Alex and Romain demonstrate how the Codex team uses GPT 5.4, the Codex Spark model, and the Codex app’s plan mode—backed by an open-source Rust harness—to one-shot generate and iterate code features like a NASA Artemis iOS screen and a 2D game at up to 1,200 edits per second.
- The Codex team writes specs in under 10 bullet points when implementing new features, relying on Codex to handle most of the coding work.
- In “fast mode” with Codex Spark, live edits to a 2D game rendered at an average throughput of 1,200 code changes per second.
- The Codex app, VS Code extension, and CLI all communicate with the same open-source Rust-based harness, allowing multiple parallel agent tasks independent of a single workspace folder.
#3 📝 Simon Willison
Eight years of wanting, three months of building with AI - Lalit Maganti details building syntaqlite: after eight years of wanting and three months of heavy AI-assisted development, they produced high-fidelity devtools for SQLite. AI helped bootstrap the prototype but struggled with high-level architecture, requiring human-led redesign for a robust final product.
#4 𝕏
Boris Cherny built custom token-counting for Bedrock, Vertex, and Azure (which lack a native API) while using Anthropic’s built-in token-counting endpoint directly.
#5 𝕏
Sebastian Raschka notes coding agents can already read markdown files in the main repo. He suggests adding a dedicated skills extension with its own folder and registry.
#6 𝕏
Logan Kilpatrick highlights that async app generation is already supported – start an app build, close the tab, and it keeps generating in the background.
#7 𝕏
Philipp Schmid released a visual guide to Gemma 4 illustrating its multimodal image/audio/text processing, sparse MoE inference that runs only a fraction of parameters, and a clever 2B-parameter embedding trick to fit on phones.
#8 in
Peter Yang shares that Alex, OpenAI’s Codex product lead, eschews medium-term roadmaps in favor of concrete eight-week sprints paired with a loose year-plus vision to guide model development.
#9 𝕏
Andrej Karpathy thinks the new Read endpoints are promising but warns that 30 minutes of hacking around cost him $200 due to steep pricing. He also criticizes the scattered short‐page docs and the lack of any mention of XMCP.
#10 in
Dharmesh Shah argues that strategy and story are one and the same—by iteratively crafting and testing a compelling, future-focused narrative that resonates with customers, you simultaneously develop and execute your strategic plan.
#11 in
Dharmesh Shah says HubSpot was the first leading CRM to launch a public MCP beta ten months ago, and now thousands of customers connect it to AI apps like Claude and ChatGPT. He asks whether opening the platform in this way makes HubSpot more valuable.
#12 ▶️
Head of Growth (Anthropic): “Claude is growing itself at this point”
Lennys Podcast
Anthropic’s growth team launches CASH (Claude Accelerates Sustainable Hypergrowth) using Claude with Opus 4.6 to fully automate growth experimentation—from opportunity identification to post-launch analysis—achieving junior PM-level win rates on copy and UI tweaks.
- Anthropic’s ARR jumped from $1 billion at the start of 2025 to $19 billion by February 2026 (10× YoY growth), hitting $4 billion mid-2025 and $9 billion end-2025—a $18 billion increase in 14 months.
- CASH was initiated a few months ago but only began delivering reliable results after upgrading from Opus 4.5 to Opus 4.6, automating four stages of growth work (opportunity ID, build, QA/brand compliance, and analysis).
- Co-work’s desktop app runs a scheduled task each morning on ~20–25 Hex chart links and Slack MCP transcripts, then uses Claude to summarize top concerns and insights in Slack.
#13 𝕏
Teresa Torres spotlights Banani’s fully autonomous design agent that pairs designer oversight with AI automation, already churning out hundreds of thousands of designs weekly.
#14 𝕏
Logan Kilpatrick notes surging demand for Gemma 4 and highlights that Google AI Edge has climbed to #8 on the iOS App Store’s productivity chart.
#15 𝕏
Andrej Karpathy argues X should make read endpoints significantly cheaper while raising write costs to rein in surging AI usage. He notes his own project was purely read-based and sees major value in exposing X’s rich data to agents in a read-only format.
#16 𝕏
Lenny Rachitsky quotes Anthropic’s Head of Growth saying exponential AI capability and diffusion tools will unlock new, ever-larger markets and boost product value 1,000× in two years.
#17 𝕏
Lenny Rachitsky highlights Anthropic’s ARR jumping from $1B to $19B in a year (adding $6B in February).