🎮️ The MMORPG for code lovers

Hey everybody!

Today’s roundup covers Anthropic’s new Claude 3.7 Sonnet and its AI coding tool, Claude Code. We also dive into OpenAI’s latest research, which finds even top AI models struggling with real-world coding tasks. Plus, we explore the power of learning in public, the rise of AI in tech interviews, and an MMORPG built for coders.

Quick Links

🆕 Claude Code & 3.7 Sonnet is here!
Claude 3.7 Sonnet, Anthropic’s most advanced model yet, offers both instant responses and step-by-step reasoning. It excels in coding and front-end development, with API users gaining fine-tuned control over its thinking time. Alongside the model, Anthropic introduced Claude Code, a command-line tool for AI-driven coding, now in limited research preview. Early reviews look pretty great.

🤖 Top AI Still Fails Most Coding Challenges, OpenAI Finds
Despite Sam Altman’s claims that AI will surpass "low-level" engineers by year’s end, OpenAI researchers found even top models fail most coding challenges. Using the new SWE-Lancer benchmark, 1,400 Upwork tasks, they tested GPT-4o, Claude 3.5 Sonnet, and OpenAI’s o1 reasoning model. Results? AI still struggles with both bug fixes and higher-level decision-making, especially without internet access.

📢 On the benefits of learning in public
Writing about what you learn helps solidify knowledge, making it stick on a deeper level. Giles Thomas’s recent blog argues that true understanding comes from explaining concepts clearly, if you can’t teach it, you don’t fully grasp it. And blogging is really powerful way to solidify what you have learnt.

🎮️ The MMORPG for code lovers
Artifacts is a Sandbox MMORPG where you can use any programming language to control your characters with their API.

💼 AI Killed The Tech Interview. Now What?
Nobody enjoys tech hiring, and AI is making traditional coding interviews even less relevant. HackerRank struggles as AI easily solves its challenges, shifting the real test to AI prompting skills. Basic comp sci questions are losing value, though architectural interviews remain safe, for now. But what happens next?

More Reading

• The TypeScript Agent Framework built by the Gatsby team.

• Google's Shift to Rust Programming Cuts Android Memory Vulnerabilities by 68%.

• How Core Git Developers Configure Git.

• Software engineering job openings hit five-year low?

The More Senior You Get, The Worse LLMs Become?

In my latest video we take a look at an article that made me rethink how I view the impact of LLMs along different levels of software developers and IT professionals.

Is it true that LLMs get more ineffective the more senior you get?

Favorite Tweets

Until next week,

Travis.