Opus 4.8 Drops; Sad Siri Set To Sing? Times Workers vs AI (Irony)

Today's AI Outlook: ⛅️

Anthropic Hits The Gas On Funding & 4.8

Anthropic released Claude Opus 4.8, pitching it as a major frontier-model jump that beats GPT-5.5 and Gemini 3.1 Pro across agentic coding, computer use, financial analysis and Humanity’s Last Exam. The launch landed alongside a reported $65B raise that would value Anthropic at $965B, putting it ahead of OpenAI by valuation, according to The Rundown AI. AI Secret took a sharper view, framing the release as a fast patch after complaints about Opus 4.7.

Why it matters

Anthropic is trying to own two stories at once: best model and biggest AI lab valuation. That is IPO catnip. But the reporting also shows the tension in frontier AI right now: users want models that are more capable, cheaper, less lazy and less hallucinatory, preferably yesterday.

The Deets

• Opus 4.8 keeps the same pricing as 4.7, at $5/$25 per million tokens, per AI Secret.
• The Rundown AI says the model leads GPT-5.5 and Gemini 3.1 Pro on most major benchmarks.
• Anthropic says 4.8 is less likely to make unverified claims and more likely to flag uncertainty.
• Fast mode is reportedly 3x cheaper.
• Claude Code is getting parallel sub-agents for complex, long-running tasks.
• AI Secret notes the launch came just 42 days after Opus 4.7 and says GPT-5.5 still leads on Terminal-Bench 2.1.
• Anthropic also teased a Mythos-class AI coming in the next few weeks.

Key takeaway

Anthropic is no longer playing lovable safety nerd. It is playing trillion-dollar contender, with all the pressure, hype and bug-fix baggage that comes with the crown.

🧩 Jargon Buster - Agentic Coding: When an AI model writes, tests and revises code through multiple steps instead of simply suggesting snippets.

Gaming Accidentally Invented An AI Menace

A Drivatar named bowie knife99 in Forza Horizon 6 has gone viral for driving like a raccoon found the nitrous button. Drivatars are AI opponents trained on real player behavior. This one rams, ambushes and flips other cars so aggressively that clips have spread across Reddit and social media, with Xbox UK reportedly posting, “happy holidays to everyone except bowie knife99.”

Why it matters

This is a tiny, chaotic preview of a much bigger AI problem: systems trained on human behavior can amplify the weirdest and worst parts of the dataset. In this case, the result is funny. In higher-stakes systems, the same pattern can get ugly fast.

The Deets

• Drivatars are built from real player driving data.
• The player behind the original behavior is unknown.
• The AI opponent appears across countless races, turning one person’s habits into everyone’s problem.
• The community backlash became memes, and the memes became accidental marketing.
• The system did what it was designed to do: make AI opponents feel human. Unfortunately, humans sometimes drive like caffeinated bumper cars.

Key takeaway

AI trained on people does not automatically produce average behavior. Sometimes it preserves the worst guy in the lobby forever.

🧩 Jargon Buster - Drivatar: A game AI opponent trained on real player driving data to mimic human-style racing behavior.

⚔️ Power Plays

Apple Wants Siri To Stop Being A Punchline

Apple’s long-delayed AI Siri revamp is reportedly taking shape, with The Rundown AI citing Bloomberg details that describe a Gemini-powered assistant, richer AI search and a ChatGPT-style Siri app. The new Siri would live inside Dynamic Island, use on-device data and screen context, and route some queries to outside AI models.

Why it matters

Apple has more than 1B iPhone users, which means even a “pretty good” Siri upgrade could instantly become one of the most-used AI products on Earth. The stakes are brutal: Apple either turns Siri into the everyday AI layer for the iPhone, or it keeps watching OpenAI, Google and Perplexity eat the interface layer.

The Deets

• Siri is reportedly being rebuilt on Google Gemini.
• Users may be able to swipe down from Dynamic Island for AI search, chat or iOS tasks.
• Answers could appear as rich cards, similar to Perplexity-style AI search.
• A deeper swipe may open a dedicated ChatGPT-style Siri app.
• Apple is also reportedly working on AI photo editing, wallpapers and natural language shortcut creation.
• The new Siri could land in the Camera app as well.

Key takeaway

Apple’s AI strategy is finally starting to look less like a press release and more like a product. The hard part is shipping it.

🧩 Jargon Buster - On-Device Data: Information stored or processed directly on your phone, such as messages, photos, app activity or screen context.

Perplexity’s Copyright Fight Gets Messy

CNN sued Perplexity AI in New York, accusing the company of copying more than 17,000 stories, videos and images. Perplexity’s response, according to AI Secret, was: “You can’t copyright facts.” CNN had reportedly tried to negotiate a license before talks failed. The Rundown AI also flagged the lawsuit in its quick hits.

Why it matters

This is the AI search fight in miniature. Publishers say AI answer engines are taking reporting, removing traffic and weakening the business model that funds journalism. AI companies say facts are free to use. The courts now get to decide where summarizing ends and substitution begins.

The Deets

• CNN alleges Perplexity copied articles and paywalled material.
• Perplexity argues facts are not copyrightable.
• CNN reportedly attempted licensing talks first.
• The fight echoes a broader publisher complaint: AI tools can answer users without sending them back to the original reporting.
• The lawsuit adds more legal heat to the AI search category.

Key takeaway

AI search is great when it saves users time. It gets legally spicy when the time it saves comes from skipping the people who did the reporting.

🧩 Jargon Buster - Copyrightable: Eligible for legal protection as original creative work, which generally covers expression but not raw facts.

The Times Faces Its Own AI Labor Fight

Unionized tech workers at The New York Times filed grievances over the company’s use of AI-related productivity tools, according to AI Secret. The tools include DX, which tracks productivity metrics, and Glean, which indexes internal documents and emails. The union says those tools are being used in disciplinary contexts without proper disclosure or bargaining.

Why it matters

This is an awkward AI mirror moment for a publisher that has aggressively challenged AI companies over consent and content use. Inside companies, the next AI labor fight may be about software judging workers through opaque metrics.

The Deets

• The union represents about 700 engineers and designers.
• DX has reportedly been cited in disciplinary conversations.
• One cited metric said a worker’s pull request count was 25% below industry standard.
• Glean indexes internal documents and emails.
• The union says management has not disclosed its AI plans clearly.

Key takeaway

AI in the workplace is moving from assistant to evaluator. Workers are going to want a look under the hood.

🧩 Jargon Buster - Productivity Tracker: Software that measures employee activity, output or workflow patterns, often using metrics like commits, tickets, pull requests or time spent in tools.

🛠️ Tools & Products

Cursor Says AI Coding Gains Are Real, But Uneven

Cursor released a Developer Habits Report showing developer output has more than doubled, according to The Rundown AI. But the boost is not evenly spread. A small group of power users is capturing huge gains, while many developers lag behind.

Why it matters

AI coding tools are moving from novelty to production infrastructure. The catch: simply handing everyone an AI agent does not make everyone a 10x engineer. Teams still need workflows, guardrails and cost discipline.

The Deets

• Lines of code added per developer per week reportedly rose from 3.6K to 8.6K in 18 months.
• Mega pull requests with 1K+ lines changed are becoming more common.
• Tool calls rose 30% in two months.
• AI-made changes reaching commits without manual review rose 5x.
• Cost per agent request varies 9x across models.
• The top 1% of developers produce 46x more code than the median active user.

Key takeaway

AI coding is boosting output, but the advantage goes to people and teams who know how to drive the machine.

🧩 Jargon Buster - Pull Request: A proposed code change that developers submit for review before it gets merged into a project.

⚡ Quick Hits

• Google doubled Omni generations for Ultra users and fixed Gemini usage-limit issues with free Flash-Lite prompts, caps on high-cost requests and better tracking.
• Glean crossed $300M in top-line revenue as AI budget control becomes a bigger enterprise selling point.
• Visa invested in Replit to support agentic payments for developers and AI-built apps.
• AWS is pushing OpenSearch toward agentic workloads as more of the web gets rebuilt for machine users.
• Rivian said AI will make the CarPlay debate obsolete as cars move toward native intelligent assistants.
• Microsoft redesigned Microsoft 365 Copilot to make AI actions faster and more central inside Office workflows.
• Foxconn Chairman Young Liu said massive AI cloud spending is fueling growth and keeping him optimistic on second-half AI server demand.

🧰 Tools Of The Day

• Founder Starter Kit: Pika’s Claude skills package helps founders move from product idea to launch workflow.
• Dubbing V2: ElevenLabs’ new dubbing AI adapts content across 90 languages, useful for creators and companies trying to make video travel without hiring a global voice cast.
• Paris 2.0: Bagel’s efficient, decentralized-trained video generation model adds another contender to the increasingly crowded video AI race.
• Computer: Perplexity’s agent is now inside Excel, Word and PowerPoint, which means office work is getting another AI copilot whether your spreadsheet asked for one or not.
• Unwrap: Pulls customer feedback from surveys, reviews, support tickets and social comments into one AI-assisted view, with customers including Perplexity, Stripe, Oura, lululemon and DoorDash.

Today’s Sources: The Internet, The Rundown AI, AI Secret