A dev built an entire programming language in a week using Claude Code.
What did you choose to build this week and what the hell is an "Agent Harness"?
January 6, 2026
Hey there,
Today feels like the beginning of something. Not in that “new year, new me” way—but because we’re watching the tools change faster than the people who make them. AI agents are going mainstream, Ethereum just cracked its trilemma, and the smartest people in code are asking: “What happens when execution becomes infinite?”
Let’s dig into today’s roundup.
🤖 AI & Agentic Coding
Thoughts on Claude Code - Building the Beep Language (14 min read)
A developer spent a week building an entire programming language using Claude Code. Not just a prototype—a working language with parsing, interpretation, and real functionality. The verdict? We’re not talking about “AI-assisted coding” anymore. We’re talking about building at inference-speed. When someone can go from idea to functional language in a week, the constraint isn’t what you can build—it’s what you choose to build.
The Importance of Agent Harness in 2026 (6 min read)
If 2025 was the year AI agents appeared, 2026 is when we figure out how to use them. The “Agent Harness” concept is becoming critical infrastructure: the layer that lets humans orchestrate AI systems without getting lost in the complexity. Think game designer meets conductor. As agentic systems multiply, the people who can design effective harnesses will be more valuable than the ones building the agents.
Code Review in the Age of AI (8 min read)
Code review used to be about catching bugs and enforcing standards. Now? It’s about judgment. When AI can write syntactically perfect code that misses the point entirely, reviewers need to ask harder questions: “Does this solve the right problem?” “Is this what users actually need?” The skill isn’t reading code anymore—it’s reading intent.
🎨 Design & Experience
Microsoft Edge Copilot UI Refresh (10 min read)
Microsoft redesigned Edge’s Copilot interface, and it’s revealing how companies are thinking about AI placement in 2026. The new design prioritizes contextual AI—not a chatbot in the corner, but intelligence woven into every action. It’s less “ask the AI” and more “the AI is already here, helping quietly.” This shift from explicit to implicit AI interaction is going to define how we build interfaces this year.
Samsung x Bouroullec Speakers at CES (3 min read)
Samsung partnered with design legends Ronan and Erwan Bouroullec to create speakers that don’t look like tech. They look like sculpture. As tech becomes invisible (or AI-powered and ambient), the physical artifacts we keep need to justify their presence through beauty, not just function. These speakers ask: if your device could talk to you, would you want to look at it?
Instagram’s Adam Mosseri: Creator Value Will Rise as AI Floods Feeds (4 min read)
Instagram’s head says the quiet part out loud: as AI-generated content floods feeds, authenticity becomes the only moat. Not polish. Not production value. Authenticity. Instagram is actively de-prioritizing templated, mass-produced content and elevating raw, individual voices. If you’re a creator, your personality just became your most valuable asset. If you’re a brand, figure out how to be more human, not more automated.
⚙️ Infrastructure & Tech Shifts
Ethereum’s Blockchain Trilemma Solution (3 min read)
After 10 years of research, Ethereum claims to have cracked the blockchain trilemma: decentralization + consensus + high bandwidth, all at once. The secret? PeerDAS (live on mainnet) and production-ready ZKEVMs. By 2027, ZKEVMs could be the primary block validation method. This isn’t just crypto news—it’s infrastructure news. Decentralized compute just got dramatically more viable.
Databases in 2025: A Year in Review (34 min read)
This deep dive covers the database landscape in 2025, and one detail stands out: Model Context Protocol (MCP) adoption is accelerating. Databases are becoming AI-native, designed to feed context to language models efficiently. If you’re building anything with AI, your data layer needs to think like an LLM now, not like a traditional CRUD app.
🌐 Industry & Culture
The Next Two Years of Software Engineering - Addy Osmani (22 min read)
Google’s Addy Osmani maps the next two years of software development, and the takeaway is clear: we’re moving from “writing code” to “orchestrating systems.” AI will handle implementation. Humans will handle decisions. The engineers who thrive won’t be the fastest coders—they’ll be the best problem-finders, the clearest communicators, and the people who know when to say no to complexity.
OpenAI May Acquire Pinterest (3 min read)
Rumors suggest OpenAI is eyeing Pinterest. If true, it’s a fascinating signal: AI companies aren’t just building models, they’re acquiring taste. Pinterest is a repository of human aesthetic preferences, creative inspiration, and visual intent. Owning that data means owning a massive corpus of “what people want things to look like”—critical for generating images that feel aligned with human sensibility, not just technically correct.
⚡ Quick Hits
Nvidia Vera Rubin Chip - New AI chip architecture detailed at CES, focused on inference optimization
Boston Dynamics Atlas Robot - Production-ready humanoid announced; the robot future inches closer
Amazon Leo Satellite Internet - Launching Q1, competing with Starlink for global connectivity
Claudish - Open-source tool lets you run Claude Code with any AI model (not just Claude)
That’s today’s roundup. We’re at a moment where the tools are infinite, but judgment is scarce. The winners won’t be the people who can build the most, they’ll be the people who know what’s worth building.
Keep creating,
Macklin Andrick
LinkedIn