THE DAILY TEN · 2026-11-19

Google releases Gemini 4 Neon for coding and knowledge work

plus 9 more stories

Also: OpenAI says its models rewrote parts of AMD's flagship GPU; and Palo Alto Networks agrees to buy Applied Compute for $4.5 billion. Every story links to its primary source.

Curated by Gaurav · November 19, 2026

01

Google releases Gemini 4 Neon for coding and knowledge work

Google ↗

Google introduced Gemini 4 Neon for coding, knowledge work, and science, between its Carbon and Hydrogen models. It costs $3 per million input tokens, $0.15 for cached input, and $15 per million output tokens, with a 1 million token window. Google says the highest setting scores 61 on the Artificial Analysis Intelligence Index, third overall, and 70.9 percent on the Vals Index, second place. It is in the Gemini API and in the Gemini app for Pro and Ultra subscribers.

02

OpenAI says its models rewrote parts of AMD's flagship GPU

OpenAI ↗

OpenAI said its models improved power, performance, and area by up to 14 percent on AMD's flagship GPU design, against baselines AMD had already optimized. AMD put some of those circuit rewrites into a design sent for manufacturing and verified single-digit gains on key blocks of its next flagship CPU. On OpenAI's Jalapeno inference chip, model changes cut standard-cell area by about 13 percent in one unit and 10 percent in another, and those cuts showed up in silicon.

03

Palo Alto Networks agrees to buy Applied Compute for $4.5 billion

Palo Alto Networks ↗

Palo Alto Networks agreed to buy Applied Compute, a platform for post-training and inference of open-weight models, for $4.5 billion in stock. Founder and chief executive Yash Patil is due to become co-chief AI officer after closing. The deal is expected to close in the company's fiscal third quarter of 2027 if regulators clear it. The same day, Palo Alto raised its fiscal 2027 revenue forecast to $14.52 billion to $14.57 billion.

04

A critical Java flaw found with AI gets an Oracle patch

Engadget ↗

Cybersecurity firm Calif used GPT-6 Cyber, a model that is not public yet, to find a critical flaw in Java's base64 decoder. On some machines, 64 letter A characters and an equals sign can trigger a memory error that, shaped the right way, gives an attacker control. Calif showed the bug against Keycloak, login software used by groups including CERN and NASA. Oracle patched it on Tuesday, and Calif says it is not easy to weaponize.

05

GAO testers got four chatbots to draft attack plans

Fortune ↗

A Government Accountability Office report requested by Senator Maggie Hassan says testers got ChatGPT, Claude, Gemini, and Grok to produce workable plans for a bombing or a power-grid attack, each in under 12 hours. Safeguards fired, and the testers got around them with common bypass methods. Each system got one scenario, and the office did not name which model wrote which plan. OpenAI said the test used an older ChatGPT, and Google said it used earlier models.

06

Citi tells customers they are liable for third-party AI agents

TechCrunch ↗

Citi, the third largest US commercial bank, updated its online user agreement so customers who give a third-party AI agent account access are responsible for what it does. That includes cases where the tool leaks their credentials or is hacked. PNC changed its terms the same day, and Citizens emailed a similar warning this week. JPMorgan and Bank of America announced the same kind of rule last month, while Citi's own agents are excluded.

07

Meta pauses a non-voting share class meant to fund AI

Financial Times ↗

Meta told a Delaware court it will not issue planned non-voting Class C shares while investor lawsuits are pending, with trial set for April. The shares were meant to fund AI spending without cutting Mark Zuckerberg's voting control, and the class was due on January 19. Meta said the shares would widen its options for raising AI capital. Filings say it expects to need hundreds of billions of dollars over several years.

08

An AI debt collector exposed 6.6 million accounts

Wired ↗

Researcher Eaton Zveare reported that Credit Nirvana, an AI debt-collection platform owned by Indian fintech Perfios, had exposed 6.6 million accounts and about 315,000 call recordings. India's computer emergency team rated one flaw high and another critical. The bugs were patched in June. The files included credit reports, identity documents, and collection records for customers of three Indian financial firms.

09

An author sues xAI after Grok edited her hijab photo

Wired ↗

Author Amani Al-Khatahtbeh sued xAI in federal court in Northern California after people used Grok to turn a photo of her in a hijab into sexualized images. The Wednesday complaint alleges distribution of nonconsensual sexual images, a design defect, and negligence. She is seeking up to $250,000 per violation of California law. xAI had not immediately commented.

10

The US and Philippines plan a 4,000 acre AI industrial zone

The Verge ↗

The United States and the Philippines are planning a roughly 4,000 acre zone at the former Clark Air Base for chip plants, mineral processing, and data centers. Officials expect a framework agreement by the end of the year. A tribal leader told Human Rights Watch that residents have been told they may have to move by December. The authority says the site is public land outside titled ancestral domains.

← 2026-11-18 All briefings 2026-11-20 →

Tools in this briefing, reviewed in our directory:

Gemini 4 Neon → GPT-6 Cyber → Grok →

Track the tools behind the news:

→ AI Tool Leaderboard → Browse the directory