The Humans in the Loop: I/O, BUILD, Anthropic Dev Day
This Week in AI for Devs: Conference Season Breakdown
Welcome to The Humans in the Loop, your executive summary of AI news for devs. I’m Andrew, the [human] author, and I’d love your feedback on this newsletter. Please feel free to email me at andrew@heavybit.com with your thoughts. Thanks!
Our top story: The Biggest News at Google I/O, MSFT BUILD, Anthropic Dev Day
AI conference season is starting in earnest with Google I/O 2025, which unveiled several announcements, the most notable being the company’s “parallel, asynchronous coding agent that works directly with your GitHub repositories,” Jules, which attempts to understand what you’re trying to code; along with GOOG’s newest top-performing reasoning model Gemini 2.5 Deep Think; plus a new version of Firebase powered by Gemini 2.5 to both improve app-building and improve integrating AI into existing apps. Microsoft’s 2025 BUILD event introduced Copilot Agent Mode, which attempts to act as a “peer programmer” that can “answer questions and automate lower-level parts of your workflow”; an open-sourced Copilot chat extension; and NLWeb, an open-source protocol (which you can try here) that attempts to facilitate building AI chat and other AI functionality directly into Web apps. And Claude creator Anthropic recently held its Developer Day event, primarily fielding media questions and reaffirming its commitment to building Claude to power software development as popular applications like Loveable continue to rely on the model, though Claude 4 users have since encountered issues (which we’ll cover below). We’ll keep you up to date on other conference announcements as they emerge.
And Now: The bottom line in AI news for busy devs.
💻 Development
Guide: Replit on Vibe Coding
Everyone’s doing it. We may as well figure out how? Replit’s guide covers how to prompt your way to an app in production.
Shopping for an MCP Server?
If you haven’t decided on which MCP server to use, data scientist Avi Chawla suggests these six open-source alternatives.
Discussion: Getting Actual Value from Coding Assistants
This lively thread covers getting value out of AI coding assistants, from green-field early stages to incremental improvements.
🤔 Interesting AI Projects, Research, and Updates
Opera Neon: “A Browser for the Agentic Web”
Opera’s new “agentic” browser not only chats with you but also executes tasks like booking flights and building apps when prompted.
[GH] OpenEvolve: OSS Response to GOOG AlphaEvolve
This new project attempts to mirror the functionality of DeepMind’s AlphaEvolve, a coding agent that continuously tackles coding challenges.
Get Startup Funding With One Weird Vibe Coding Stack
Peek.money founder Sherry Jiang says she used GPT, v0, Cursor, Supabase, and Vercel to nab $275K in accelerator funding.
💼 Hiring and Community
Startups Hiring This Week:
- Director, Field Engineering → Together AI
- Senior Research Scientist → Flawless AI
Mid-Markets Hiring This Week:
- AI Agent Engineer → Farther Finance
- AI/ML Engineer → AlertMedia
Enterprises Hiring This Week:
- AI Engineer → Capco
- Sr. Data Science Analyst → ORCL
In the Talent-Starved AI Market, Is Retention the New ROI?
Reports suggest that META has found it challenging to retain top research talent, while Anthropic has apparently retained nearly 80% of new hires.
💡 Spotlight: Newly-Launched AI Startups
- LMArena: $100M from a16z to [continue to] rank AI model performance
- Traceloop: $6.1M from Sorenson Capital for managing AI agents
- Bito: $5.7M from Vela Partners for AI-powered code review
🚀 New Model Launches
- Devstral: Mistral’s new coding model with a larger context window
- DeepSeek R1: Re-launched at 685B parameters
- AceReason Nemotron-14B: NVDA’s 100% RL-trained math-and-code model
- Claude 4: “The world’s best coding model” (but not without issues, see below)
🏭 Industry: M&A, Launches, Trends
Claude 4’s Interesting Launch
While the latest Claude gets high marks for coding and reasoning, its apparent tendency to call the cops on users it deems ‘immoral’ and its apparent sec vulnerabilities with GitHub MCP are less encouraging.
OpenAI Buys io for $6.5B, Stargate UAE & Texas, o3 Won’t Shut Down
The GPT vendor has acquired AAPL designer Jony Ive’s company io to create a commercial device of unspecified design while inking Stargate-related commitments to data centers in Abu Dhabi and Abilene. (Also, its latest reasoning model doesn’t seem to like it if you tell it to shut down.)
META Launches “Llama for Startups” Program
META launches a startup program offering extra model support and potential financial support for new orgs that adopt Llama models.
⚖️ Copyright, IP, Licensing, and Regulation
Federal Budget Bill Has 10-Year AI Regulation Ban Rider
The much-discussed US budget bill had the House’s proposed 10-year ban on any regulation of AI slipped in at the 11th hour.
Clegg: Asking Artist Permission Would “Kill the AI Industry”
Former UK deputy PM Nick Clegg suggests that AI companies having to attain permission from artists before training models on their materials would “basically kill the AI industry in this country overnight.”
About the Author
Hi there. My name is Andrew, and I work at Heavybit, the leading VC for software infrastructure startups. As the Editorial Lead, my goal is to find the most valuable and important AI news for developers and founders. The idea is to curate and bottom-line emerging trends from our perspective in 10+ years of coaching and funding enterprise infrastructure companies. Email me and let me know what you think.
That’s all for this edition. In the meantime, please feel free to follow The Humans in the Loop on LinkedIn, BlueSky, and Twitter.