In partnership with

☀️ TRENDING AI NEWS

🤖 GPT-6.2: OpenAI released Astra and Sol, with Astra's API price down 20% to $8/$40 per million tokens.

🏢 Nvidia: The board says it does not intend to reinstate Jensen Huang as revenue roughly doubles.

🛠️ Claude Code: Code Mode lets Claude chain Bash, Read and Edit calls in one JavaScript step.

🚨 Congress: Huang will testify before several committees on December 1.

Nvidia just doubled its revenue and the board still told everyone, on the earnings call, that the founder is not coming back. Strange day for the world's most valuable chip company, and it was not even the only big AI story.

🛠️ Tool of the day
 

Elicit - an AI research assistant that searches academic papers and writes structured reports with citations. It helps students, analysts and anyone who needs evidence fast. It has a free plan with limited usage.

🤓 AI Trivia
 

Perplexity's new Multiagent mode runs a lead agent that coordinates a team of other agents. What is the maximum number of agents it can run at once?

Login or Subscribe to participate

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

🤖 OpenAI Ships GPT-6.2 Astra and Sol
 
Server rack with a falling price dial and two glowing orbs

A 20% price cut on the flagship

OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17. The headline for developers is price: Astra's API cost drops 20% to $8 input and $40 output per million tokens.

Astra also ranks second on the Artificial Analysis Intelligence Index, so you are not trading quality for the cheaper rate. If you want to see how the new pricing changes your own bill, our token calculator is a quick way to model it.

Run your numbers in the Token Calculator, or Read the full story →

The bottom line

Frontier pricing keeps sliding. If you built a product around last quarter's API costs, your margins just improved without you touching a line of code.

You know the problem. Now build the fix.

You've spent years close to a broken industry or process. You know exactly what would fix it. But you’ve never had the time or resources to make it happen.

With Lovable, your expertise is all you need. Describe the solution to your problem — a website, a dashboard, an internal tool — and get professional-grade software back. You make every call on how it works, because you're the one who knows what it needs to do.

Start today and take $5 off your first month.

Discount of $5 applies to first month of any paid plan. You need to sign up via this webpage in order to claim the discount. The discount applies to new and returning users without an active paid plan. Terms apply.

🏢 Nvidia Doubles Revenue, Board Shuts the Door on Huang
 
Boardroom table with an empty chair and a doubling chart above a chip

Record numbers, awkward earnings call

Nvidia reported its first quarterly earnings since the ouster of CEO Jensen Huang. Revenue beat expectations and roughly doubled year over year, but interim CEO Colette Kress initially skipped any mention of Huang or the circumstances of his departure.

Board chair Harvey Jones then opened the call by saying the board "does not intend to reinstate Huang as CEO," and that the company would not be addressing the matter further. That lands against a backdrop of Huang saying he would be "honored" to return, a petition with well over 100,000 signatures, and public backing from Elon Musk and President Trump.

Huang is scheduled to testify before several congressional committees on December 1, alongside Nvidia's general counsel, over what lawmakers call a potential breach of national security. Separately, the Justice and Commerce Departments say they are investigating Megaspeed, the Singapore data center company tied to the scandal.

The bottom line

The business is booming while the leadership story is unresolved. Expect the December 1 hearing, not the earnings, to drive the next round of headlines.

🛠️ Claude Code Learns to Write JavaScript Mid-Task
 
Terminal with tool blocks chained into a single step

Code Mode chains tool calls in one step

Anthropic added an experimental Code Mode to Claude Code v2.1.323. Instead of calling Bash, Read and Edit one at a time, Claude can now write JavaScript that chains those calls together in a single step.

It also keeps talking while long-running commands execute, so you are not staring at a frozen terminal waiting for a build to finish. For anyone using

coding agents daily, fewer round trips between model and tools means faster, cheaper sessions.

The bottom line

This is a small flag that points at a big trend: agents are being given real programming constructs, not just a menu of single actions.

🤖 Perplexity Puts 16 Agents on One Problem
 
A lead agent coordinating a ring of small agent orbs

Multiagent mode cross-checks its own work

Perplexity added Multiagent to Computer, an effort level above Ultra. A lead agent coordinates up to 16 agents running on Claude, GPT and Gemini models, which share findings and check each other's work.

Meanwhile, SID AI took a similar swarm approach to search. Its SID-2 model scores 0.585 NDCG as a single agent in 3.11 seconds, and a 12-agent swarm hits 0.720 NDCG, which SID says beats GPT-6 Astra on its own benchmark.

The bottom line

More agents mostly means better answers, at the cost of more compute. The interesting question is when that trade stops being worth it.

🛠️ Google's Spark Agent Becomes Free 'Tasks'
 
Phone with floating calendar, meal plan and email cards

Google rebranded its Spark AI agent as Tasks and made it free. It runs on the new Gemini 4 Hydrogen model and can build meal plans, manage your calendar, and dig through email for recurring expenses you may not need.

You switch from the Chat toggle to Tasks, and US users get access first. Making a capable agent free puts real pressure on paid assistants.

💬 Gen Z, AI Friends and the Word 'Human'
 
A young person alone with a phone and a faint chat figure beside them

A culture check from two different angles

A new Gallup report finds that over a third of Gen Z-ers rely on AI for friendship, and that young people who use AI as a social companion or have online-only friends face higher odds of loneliness than those who do not.

On the same theme, Cambridge dictionary named "human" its word of the year, citing how the meaning and use of the word has grown more important in the age of AI chatbots, three years after ChatGPT arrived.

The bottom line

The tech is getting better at imitating company. The data suggests it is not a substitute for it.

Quick plug: if you need a website fast, our sponsor 60sec.site builds one with AI in about a minute. And visit dailyinference.com for our daily AI newsletter.

🌎 Trivia Reveal
 

The answer is 16! Perplexity's Multiagent mode lets a lead agent coordinate up to 16 agents across Claude, GPT and Gemini models, all sharing findings and checking each other's work.

💬 Quick Question
 

With agents getting cheaper and more capable, would you hand a real task like your calendar or inbox to a free AI agent? Hit reply and tell me where you draw the line. I read every response!

That's it for today - see you tomorrow with more.

🎁 Share Daily Inference
 

Share Daily Inference with a friend. You get the AI Tools Starter Kit at 1 referral and the AI Insider Briefing at 5 referrals.