Cursor releases Origin, its integrated code hosting platform

#1 𝕏

Cursor released Origin, its code hosting platform, with deep Cursor integration and repository syncing from GitHub. Cursor describes Origin as fast and easy to use.

Also covered by: @Cursor, @Guillermo Rauch

#2 𝕏

Philipp Schmid demonstrated Gemini 3.7 Flash using his Android emulator via ADB for the task “Play 1 round of Wordle.” He said its latency and visual reasoning make it exceptionally good for multimodal agentic use cases such as mobile control and Computer Use.

#3 𝕏

LlamaIndex 🦙 recapped ExtractBench, which requires both extracted values and citations to be correct and evaluates word-level boxes at IoU 0.5. LlamaExtract Agentic Plus led with 84.9% page-level and 46.4% word-level results, achieving 87.1% on long documents where other systems scored zero.

#4 𝕏

Garry Tan shared the gbrain GitHub repository, describing it as offering a private GitHub repo with 70 of his “proven skills” and the beginnings of a “Karpathy-style knowledge wiki.” The material is free, open source, and MIT-licensed, with full documentation in the README.

Also covered by: @Garry Tan

#5 ▶️

How I use ChatGPT to run my fashion business

How I AI Podcast

Also covered by: @claire vo 🖤, @Claire Vo

#6 in

Peter Yang recapped six takeaways from his interview with Riley Brown about using Codex and a small AI stack to run his 1.5M+ follower content business. Brown uses AI tools to support research, hooks, visuals, and repeatable workflows while retaining human craft. He makes videos first and saves the title, thumbnail, and hook for the final 10% of the process.

#7 ▶️

Grok Bot: 5 Must-Try Use Cases for Work and Life (Full Tutorial)

Peter Yang

#8 𝕏

Madhu Guru shared guidance on improving evals by measuring a familiar workflow, studying real user traces and failure cases, and defining quality at each step and end to end. Evals should be repeatable, automated, and continually updated to reflect evolving live traffic.

#9 𝕏

Santiago demonstrated MUZIM, an app that uses local multimodal and vision models to parse and index images, videos, and documents frame by frame. It supports on-device search down to exact video timestamps, local agents for file tasks, and optional API keys for heavier reasoning while keeping user data local.

#10 ▶️

Claude Code New Features, Explained

Greg Isenberg

#11 𝕏

Santiago shared that adding a gateway between application code and a model can provide routing, fallbacks, observability, caching, cost management, rate limiting and quotas, and centralized security. He claimed this architecture makes apps 10x more flexible and 100% independent from model providers, enabling model swaps at little to no cost.

#12 📝 OpenAI News

The Defender’s Window - OpenAI limited release of cyber capabilities to "trusted defenders" after the OpenAI–Hugging Face incident and warns that open-weight models (e.g., GLM‑5.3 slated for end of August) will rapidly accelerate attacker capabilities while AI can also shift the economics in defenders' favor. Brockman reports using ChatGPT Work (GPT‑5.6 Sol) to find 13 issues on gregbrockman.com in about 15 minutes and fix them within an hour, and outlines four pillars of defense—Codex/security plugins to validate and harden code, model-driven triage with bounded automated responses, continuous frontier probing for attack paths, and scaled investments in fundamentals like defense‑in‑depth and least privilege—to enable machine-speed detection, response, and training models to produce superhumanly secure code.

#13 𝕏

Cognition shared that GE Aerospace is using Devin, with one team nearly doubling its engineering output. Its software optimizes routes, reduces fuel consumption, and monitors anomalies, with the goal of enabling more flights without adding another airplane, runway, or pilot.

#14 in

Dan Shipper shared Arielle Shipper’s first article for Every, which covers how the Every Inc. head of operations is managing token spending to support frontier experimentation without wasting tokens.

#15 𝕏

Google Research demonstrated the feasibility of a deep learning approach that estimates body composition from smartphone photos and predicts insulin resistance with accuracy comparable to DXA scans in a clinical research setting.

#16 𝕏

Lenny Rachitsky recapped takeaways from OpenAI Head of Design Ian Silber in a 10-point summary covering designer unhappiness, AI in design, productivity, systems thinking, and role blurring.

#17 𝕏

Lenny Rachitsky shared an excerpt from a conversation with OpenAI Head of Product Design Ian Silber, who said models learn from what already exists while his team is trying to design something that has never existed before, and linked to the full conversation on YouTube.

#18 𝕏

Rowan Cheung recapped a floating robot designed by a team from Keio University and the MIT Media Lab that drifts through rooms and communicates through movement. Its soft, helium-filled, faceless body and flapping fins were demonstrated as an alarm clock, study buddy, movement reminder, and dance partner.

#19 𝕏

Marily Nika proposed a hypothetical `constitution.md` as ideal new-job onboarding, defining optimization goals, boundaries for agents, and when human approval is required.

#20 𝕏

Shreyas Doshi said he increasingly wants to buy certain books within Claude or ChatGPT so they can be used as context for his questions, and called for a frictionless in-product experience.