OpenAI Just Cut GPT-5.6 Luna by 80%

IN THIS ISSUE

Welcome back to The AI Field.

AI is getting cheaper, more useful, and harder to trust by default. Today’s edition shows where automation now earns its place and where human judgment is becoming more valuable.

- OpenAI cuts Luna’s price by 80%, making background agents far cheaper to run
- Google uses AI to fix 1,072 Chrome security bugs across two releases
- Snapchat stops rewarding videos made entirely by AI

Read time: 4–5 minutes

Nobody became a developer to write documentation. But the docs still need to get written — PRDs, README updates, architecture decisions, onboarding guides.

Wispr Flow lets you talk through it instead. Speak naturally about what the code does, how it works, and why you built it that way. Flow formats everything into clean, professional text you can paste into Notion, Confluence, or GitHub.

Used by engineering teams at OpenAI, Vercel, and Clay. 89% of messages sent with zero edits. Works system-wide on Mac, Windows, and iPhone.

Try Wispr Flow free

RECENT UPDATES

01 · OPENAI

OpenAI cut GPT-5.6 Luna’s API price by 80% on July 30. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens. Terra fell 20% to $2 for input and $12 for output.

OpenAI also added a Fast mode for Sol that runs up to 2.5 times faster at twice the standard price. The company says Sol helped rewrite production kernels, cutting serving cost by 20% and improving token-generation efficiency by more than 15%.

OpenAI says Luna now matches work that needed frontier-class models a year ago at roughly six cents on the dollar per task and nearly nine times the speed. That is the company benchmark, not a neutral evaluation, but the price change is real and available now.

The practical move is model routing. Keep the hardest planning and judgment calls on a stronger model. Move repeatable, low-risk steps to Luna only after a small evaluation proves the quality holds.

The AI Field: The model race is shifting from benchmark wins to completed work per dollar. The winner will be the model a company can afford to call 10,000 times, not the one it saves for its ten hardest prompts.

Independent check: Axios confirms the new prices and timing

TRY THIS TODAY

Give AI your goal and constraints, then make the final decision yourself.

1. Open ChatGPT, Claude, Gemini, or your preferred AI tool.

2. Paste this prompt:

3. Correct anything the AI misunderstood. Then start the first concrete action.

02 · GOOGLE

Chrome 149 and 150 fixed more security bugs than the previous 23 releases combined. AI agents now find bugs, triage reports, propose several fixes, review each other’s work, and write tests before a developer approves the result.

One agent found a sandbox escape that had survived for more than 13 years. Google runs these agents on locked-down machines without general internet access and blocks access outside approved code folders.

The old triage process took 5 to 30 minutes for each report. Google estimates the automated version now saves hundreds of developer hours every month. In May alone, its systems also blocked more than 20 vulnerabilities from reaching production.

The AI Field: This is better proof than another app built from one prompt. Builders should copy the chain, not just the tool: finder, fixer, critic, tests, then human approval.

Independent check: TechCrunch reviews Google’s reported results

03 · CREATOR ECONOMY

Wholly AI-generated videos are no longer eligible for Spotlight recommendations. AI-assisted and AI-edited work can still qualify, including content made with Snapchat’s own tools, but it will carry transparency indicators.

This is not a platform-wide ban. It is a distribution decision about what earns free reach.

Snap says the number of unique Spotlight contributors grew more than 120% from a year ago. The platform has enough human supply to choose what it rewards, and it is choosing work where AI supports the creator instead of replacing one.

The AI Field: Creators just got a clear signal. AI can speed up production, but it cannot be the whole reason to watch. When generation is cheap, taste and a real point of view become scarce.

Independent check: TechCrunch explains the recommendation change

THE AI TOOLBOX

FEATURED TOOL

What it does: Uses Chrome, logged-in accounts, and saved passwords to work through a web task with your permission. Sensitive steps such as payments return to you.

Best for: Comparing travel options, collecting research, and completing repetitive browser steps.

Pricing: Spark access is expanding to Google AI Pro subscribers. Chrome auto browse is initially rolling out in the US.

Worth trying: Give it one low-risk research errand and keep the final decision manual.

Also on our radar:

QUICK HITS

AI IN PRACTICE

REMOTE WORKFLOW

Alex uses one always-on Mac as the home base for his Codex work. His other devices connect to it so he can work from anywhere.

The simple version:

Advanced setup: Alex adds Tailscale for private access and Hermes Agent for longer tasks. Neither is required for Codex Remote.

The AI Field: Start with one trusted workspace. Add the extra layer only when you need more control.

YOUR TURN

Until next time!
Olle
Founder of The AI Field