Welcome back to Daily Zaps, your regularly-scheduled dose of AI news ⚡️

Here’s what we got for ya today:

Let’s get right into it!

Sign Up | Advertise | Tools

OpenAI slashed pricing across its GPT-5.6 lineup. Luna, the workhorse model, now costs $0.20 per million input tokens and $1.20 per million output tokens — an 80% cut. Terra, the mid-tier model, dropped 20% to $2 and $12 per million tokens, while Sol Fast, the low-latency option, is now up to 2.5x faster than its predecessor at roughly 2x the price. On Artificial Analysis' Intelligence Index v4.1, Luna sits near the top of the frontier curve while undercutting comparably capable models like Claude Opus 5 Low and GLM-5.2 Max by a wide margin on cost per task.

The cuts reset the calculus for teams building on GPT-5.6: what matters isn't raw intelligence score but outcome-per-dollar — how much reliable output a workflow gets per unit spend. At Luna's new price, high-volume steps like classification, extraction, and drafting become close to free to run at scale, which makes tiered routing the more attractive architecture: send bulk work to Luna and reserve Terra or a frontier model for the smaller share of steps — ambiguous judgment calls, safety-sensitive decisions, high-stakes approvals — where errors are costly. Sol Fast's speed gain matters here too, since latency-sensitive routing paths can now clear more steps per second without changing which model carries the risk.

Read more → OpenAI

OpenAI analyzed more than 800,000 work-related messages from U.S. ChatGPT users and found that 16.8% of all work messages — and 43.5% of occupation-specific messages — involved tasks associated with another occupation entirely. Examples ranged from marketers troubleshooting websites to salespeople analyzing customer data, suggesting day-to-day work increasingly spills across traditional job boundaries rather than staying confined to a single role's core tasks.

The pattern points to AI reducing the need for specialist handoffs and quietly reorganizing how roles are structured, an effect OpenAI says is most pronounced at small businesses that can't staff a dedicated specialist for every function. Still, the data measures observed ChatGPT usage patterns, not proof of economy-wide job outcomes — it shows what people are asking AI to help with, not how headcount, wages, or role definitions are ultimately changing across the labor market.

Read more → OpenAI

Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

Google DeepMind's Gemini Robotics 2 splits robot intelligence into a three-model stack: a vision-language planner that reasons about the task and its environment, a whole-body controller that coordinates torso, arms, and legs as one system rather than separate subsystems, and a dexterity model tuned for fine manipulation. In demos, robots handle shelf-stocking, chess moves, and tray-clearing, and multiple robot bodies coordinate on a shared task — an early step toward fleets that divide labor instead of one robot working in isolation.

The bigger claim is transfer: DeepMind says the stack adapted to an entirely new robot body — different arm geometry and sensors — using fewer than 200 demonstration examples, far below what earlier generations needed to reach comparable competence. That points toward a general robotics "brain" that doesn't need retraining from scratch per hardware platform. The gap that remains is dexterity: multi-finger tasks, like precise pinches and tool handoffs requiring independent finger control, stayed uneven across trials — the clearest sign that whole-body gains haven't fully carried over to fine manipulation yet.

Read more → Google DeepMind

Microsoft and Meta reported sharply rising AI infrastructure costs. Microsoft’s quarterly capital spending jumped 70% to $41 billion, with roughly two-thirds going to short-lived CPUs and GPUs; Meta’s expenses rose 55% to $42 billion as it lifted the low end of its 2026 capex forecast to $130 billion.

Microsoft paired the spending with $90 billion in quarterly revenue and 43% Azure growth, while Meta’s free cash flow fell 91% year over year. The divergence is becoming the central AI-business test: companies are all buying compute at historic scale, but investors increasingly want proof that revenue and margins can keep pace with the replacement cycle.

Read more → Axios

In case you’re interested — we’ve got hundreds of cool AI tools listed over at the Daily Zaps Tool Hub.

If you have any cool tools to share, feel free to submit them or get in touch with us by replying to this email.