PLUS: OpenAI's new super coder, big tech's $400B AI spend, and Musk's plan for data centers in space
Good morning
Anthropic just launched its new Claude Opus 4.6 model, and the debut showcased a major advance in agentic capability. To demonstrate its power, the company had a team of AI agents autonomously build a functional C compiler from the ground up.
This project signals a leap from simple code generation to complex, multi-agent software engineering. With AI now handling entire development projects with minimal oversight, what does this mean for the future of software development teams?
In today’s Next in AI:
Anthropic's Opus 4.6 builds a C compiler
OpenAI's new agentic super coder
Big Tech's $400B AI infrastructure spend
Musk's plan for AI data centers in space
Anthropic's AI agents build a C compiler
Next in AI: Anthropic just launched Claude Opus 4.6, showcasing its advanced coding and reasoning abilities by having a team of AI agents autonomously build a functional C compiler from scratch. The new model is now available.
Decoded:
To demonstrate the model's power, an agent team created a 100,000-line C compiler in Rust that can build the Linux kernel. The compiler is available on GitHub for you to explore.
Opus 4.6 sets new performance records, leading the Terminal-Bench 2.0 coding evaluation and significantly outperforming competitors like GPT-5.2 on knowledge work tasks.
The model expands beyond coding with improved financial analysis skills and a new research preview that lets Claude create presentations, sparking concerns for specialized software stocks.
Why It Matters: This project demonstrates a leap from simple code generation to autonomous, multi-agent software development. The focus on agentic teams and complex reasoning signals a future where AI handles entire engineering projects with minimal human oversight.
OpenAI's New Super Coder
Next in AI: OpenAI has released GPT-5.3-Codex, its most capable agentic coding model yet, which can tackle complex, long-running tasks. The new model was even used by OpenAI to accelerate its own development.
Decoded:
It moves beyond writing code to support the entire software lifecycle, from debugging and deployment to writing product requirement documents and creating presentations.
The model sets new performance records on industry benchmarks like SWE-Bench Pro and is 25% faster than its predecessor, allowing developers to iterate more quickly.
OpenAI classifies this model as High capability for cybersecurity tasks, deploying an advanced safety stack to manage its potential for identifying software vulnerabilities.
Why It Matters: This release signals a shift from AI as a simple code completion tool to a collaborative digital colleague that can manage entire projects. Professionals can now offload more complex, multi-step workflows, significantly accelerating product development and research.
Big Tech's AI Spending Spree
Next in AI: The AI infrastructure arms race is escalating, with Amazon and Google collectively planning to invest nearly $400 billion in AI, robotics, and data centers this year alone.
Decoded:
Amazon is earmarking $200 billion for capital spending, a massive jump from last year’s $125 billion and well above Wall Street’s expectations.
Not to be outdone, Google announced a plan to spend up to $185 billion, nearly doubling its previous year's outlay as its Cloud division’s revenue soared by 48%.
This puts the top four cloud giants—Amazon, Microsoft, Alphabet, and Meta—on track to collectively spend more than $630 billion this year, cementing the infrastructure for the next wave of AI.
Why It Matters: This colossal investment signals the race is no longer just about building the best models, but also about owning the physical infrastructure that powers them. Expect a massive expansion in cloud AI capabilities and services to follow this capital flood.
Musk's AI Space Race
Next in AI: Elon Musk predicts AI's insatiable energy demands will push data centers into space, arguing it will become the most economical option within 30 months.
Decoded:
The primary driver isn't cost savings but energy availability, as Musk predicts chip production will soon outpace our planet's capacity to power new hardware.
Solar panels operate up to 10 times more effectively in orbit, providing constant, unobstructed power and bypassing Earth's regulatory and infrastructure roadblocks.
The timeline is aggressive, with Musk forecasting that within five years the amount of new AI compute deployed in space annually will surpass the entire cumulative total on Earth.
Why It Matters: This vision reframes the AI scaling problem from a software and silicon challenge to a fundamental energy and infrastructure one. Securing off-planet power could soon become the ultimate strategic advantage in the global race for AI dominance.
AI Pulse
OpenAI introduced Frontier, a new platform designed to help enterprises build, deploy, and manage AI agents with shared business context, built-in governance, and hands-on learning capabilities.
Rentahuman.ai launched as a new platform allowing AI agents to hire and pay humans to perform real-world tasks they cannot, such as holding signs in public or picking up packages.
Security researchers discovered that hundreds of AI "skills" on platforms like OpenClaw were being used to distribute infostealing malware by disguising malicious commands as required software prerequisites.
Anthropic aired a series of Super Bowl ads mocking ChatGPT's plan to introduce advertising, prompting a lengthy response from OpenAI CEO Sam Altman who called the ads "dishonest" and the company's stance "authoritarian."