The Agent Stack #018 — Friday Signal

The AI industry just showed its hand. OpenAI launched a £80/month Pro tier this week, filling the massive gap between Plus (£16) and Enterprise (£160). This isn’t about serving power users better. It’s about desperately finding revenue before the money runs out.

The profitability cliff is real

OpenAI’s new pricing tier reveals something uncomfortable: the current economics don’t work. The company burned through billions last year whilst charging £16/month for unlimited access to frontier models. That’s unsustainable when training runs cost hundreds of millions.

The £80 Pro plan offers “5x more usage” of Codex and longer sessions. Translation: they’re rationing compute for everyone else. The freemium model that hooked millions of developers is being quietly strangled.

This comes as Florida launches a criminal investigation into OpenAI over a shooting allegedly planned using ChatGPT. The company faces lawsuits and regulatory pressure whilst trying to prove it can actually turn a profit. Not ideal timing for aggressive price increases.

Meanwhile, Meta dropped Muse Spark this week and shot to #5 on the App Store. Free, capable, and backed by advertising revenue that actually works. That’s the competition OpenAI faces whilst trying to 5x their revenue per user.

The takeaway: If you’re building agents that depend on OpenAI APIs, budget for much higher costs in 2026. The race to profitability means the party’s ending.

Quick hits

• Sierra’s Ghostwriter launched as an “agent that builds agents” service. Bret Taylor claims we’re moving past clicking buttons entirely. The demo looked impressive but pricing wasn’t announced.

• Google and Intel deepened their AI infrastructure partnership to co-develop custom chips. Demand for compute is driving a global shortage, pushing more companies to build their own silicon.

• Anthropic limited Mythos model release citing cybersecurity concerns. The model apparently finds security exploits too well. Classic dual-use problem hitting frontier labs.

One thing to try

Test your agent costs with OpenAI’s new Pro tier pricing in mind. If you’re using GPT-4 heavily, the £80/month consumer plan might actually be cheaper than API calls. Run the numbers before OpenAI forces the decision.

The agent economy is repricing itself in real time. Build accordingly.