AI News Roundup – June 5, 2024

1. Recursive Superintelligence (RSI) funding – After leaving Meta, former FAIR research director Tian Yuandong (田渊栋) joined the newly announced AI lab RSI, founded by eight top AI researchers. The lab disclosed a $650 million financing round that values it at $4.65 billion.

2. Anthropic’s engineering team approach – Fiona Fung, engineering director of the Claude Code team, wrote about how their AI‑era engineering team operates. She describes a shift of bottlenecks: historically, software‑engineering processes (waterfall, agile, review meetings) aimed to reduce the high cost of writing code and engineers’ time by planning, documenting, and aligning early.

3. Odoo’s open‑source ERP impact – The long‑standing open‑source ERP Odoo continues to erode market share from heavyweight suites such as SAP and Oracle. It offers a modular platform that covers CRM, inventory, finance, and HR for small‑to‑medium businesses, and its GitHub repository highlights its focus on becoming “enterprise‑grade open‑source infrastructure.”

4. Cursor design‑engineer hiring – Cursor announced a design‑engineer position. The author, an 8‑year veteran at ByteDance who oscillates between motion design and design engineering, notes that the Doubao mobile team seeks candidates with Android platform expertise, a focus that differs from the web‑centric design‑engineering roles common in Silicon Valley.

5. AI‑powered English‑learning guide – A GitHub‑hosted English‑learning guide (over 51 k stars) rejects traditional methods such as rote memorisation, endless drills, or hoarding resources. Instead, it combines AI with scientific learning techniques to treat English as an acquireable skill.

6. Alternative English‑learning narrative – Another guide reflects on how people allocate time to meet, like, love, and eventually forget others, using this emotional timeline as a metaphor for language learning.

7. OpenAI Codex Sites – Codex Sites lets users input work content, ideas, or plans and automatically generates an interactive website or app that can be shared via a URL. This compresses the traditional prototype workflow (requirements → wireframes → engineering → demo) into a single AI‑driven step.

8. Code‑base understanding with “Understand Anything” – The tool scans large repositories into an interactive knowledge graph. It uses Tree‑sitter to extract functions, classes, and dependencies, while an LLM adds semantic explanations. Users can explore node relationships, search business keywords, and view diff‑impact analysis to see which modules are affected before modifying code.