Trending Thursday #35

Hi !

This week in Silicon Valley, startups & tech:

• AI model wars escalate: Anthropic drops Opus 4.6 with "agent teams" and 1M context, OpenAI counters with GPT-5.3-Codex

• Agent infrastructure explodes: OpenAI Frontier, GitHub Agent HQ, and Apple Xcode all launch agent orchestration platforms this week

• ElevenLabs hits $11B: Voice AI unicorn raises $500M, reports $330M ARR, so voice is the next interface war

• Software "extinction" panic: Stocks crash (Adobe -7%, Salesforce -7%) but insiders call it "nonsense"

• Product launches galore: v0 2.0, Grok Imagine 1.0, skills.sh, but really it’s all about how the Moltbook story is still a thing

As we continue to build at Context Ventures, we've launched Startup Intros with a simple mission: to help early-stage founders find the right investors, faster and smarter.

🤖 The Agent Infrastructure Arms Race Continues

Every major platform launched agent infrastructure this week. The battle isn't "who has the best model" anymore; it's "who controls the agent orchestration layer."

🟢 OpenAI: Frontier + Codex: The Full Stack Play

OpenAI dropped two major releases this week, and together they tell a clear story: they're not just building models anymore; they're building the infrastructure to deploy them.

OpenAI Frontier is their new platform for managing AI agents at enterprise scale. Shared context, onboarding, permission boundaries, basically the "AWS for agents." It's currently limited to select customers, but this is OpenAI's answer to the question every enterprise is asking: how do we actually deploy agents at scale?

GPT-5.3-Codex goes beyond coding to "nearly anything developers and professionals can do on a computer." It runs 25% faster, handles longer-running tasks, and here's the wild part: OpenAI says it's "the first model instrumental in creating itself." This isn't an incremental upgrade. It's a category shift from "coding assistant" to "universal agent."

The Play: OpenAI is building both the brain (Codex) and the nervous system (Frontier). They're not selling intelligence anymore; they're selling infrastructure.

🟣 Anthropic: Opus 4.6 + Ad-Free: The Trust Play

Anthropic dropped Claude Opus 4.6 this week, and it's not just an upgrade; it's a statement about where AI agents are heading.

• Agent Teams: Opus 4.6 can now split larger tasks into segmented jobs, coordinating multiple agents to tackle complex workflows.

• 1M Context Window (beta):4x the previous context in memory. For enterprise use cases, this is massive. Claude can now hold

• The Security Flex: During testing, Opus 4.6 found 500+ previously unknown high-severity security flaws in open-source libraries with minimal prompting.

• 90.2% on BigLaw Bench: The highest score for any Claude model, plus direct PowerPoint integration and 300K+ business users.

The Trust Play: Anthropic doubled down, saying that "Claude will remain ad-free." No sponsored links, no advertiser influence on responses. In a world where AI becomes your second brain, that's a positioning play against OpenAI's potential ad model.

The Play: Anthropic is betting on security and trust. While competitors explore monetization through ads and partnerships, Anthropic is saying, "We won't manipulate your AI assistant." That matters.

🔵 Microsoft/GitHub: Agent HQ: The Dev Play

Microsoft's playbook hasn't changed: someone else innovates, Microsoft integrates.

GitHub Agent HQ now integrates Claude and Codex agents directly into GitHub, GitHub Mobile, and VS Code. Available to Copilot Pro+ and Enterprise users, developers can run Anthropic's Claude alongside OpenAI's Codex and Microsoft's Copilot, all within GitHub's ecosystem.

Sounds Smart: Multi-agent support, not locked to one provider, pick the best tool for the job. In practice, this is Microsoft doing what Microsoft always does, showing up late with a platform play after the real builders (Anthropic, OpenAI) have already shipped the products people actually want.

The Pattern: Cortana was going to beat Siri. Bing was going to beat Google. Copilot was going to own AI coding. Now they're building an "orchestration layer,” which is what you call it when you can't win on product, so you try to win on distribution.

Enterprise Reality: Microsoft isn't building the best agent. They're building the outdated waiting room where other people's agents sit.

🏈 This Sunday: Brian's Super Bowl Tailgate (SF)

Ramp & Vently are throwing a 1,000-person Super Bowl tailgate at Fort Mason featuring Brian Baumgartner (Kevin from The Office).

The deal: dress up as Brian for a chance to win Super Bowl tickets. Brian himself is judging. Suits strongly encouraged. Free bald caps at brianshair.com.

📅 Sunday, Feb 8 from 10:00 AM - 12:30 PM
📍 Festival Pavilion, Fort Mason, SF

RSVP here!

🎙️ Voice AI Goes Unicorn

Three stories this week point to the same thing: voice is becoming the next platform war.

ElevenLabs: $11B and Counting

ElevenLabs just raised $500M at an $11B valuation, bringing total funding to roughly $800M. Sequoia led.

The number that matters: $330M in ARR. For a company that didn't exist three years ago, that's an absurd growth trajectory. ElevenLabs has become the default voice infrastructure for everyone from Hollywood studios to indie game devs to the guy who needs his AI agent to sound like it went to boarding school.

What they're building goes beyond text-to-speech. Voice cloning, real-time dubbing, conversational AI: the thesis is that every piece of software eventually needs a voice layer, and ElevenLabs wants to be the API you call.

Mistral Voxtral Transcribe 2

Mistral quietly dropped Voxtral Transcribe 2: a new speech-to-text family with speaker diarization and ultra-low latency. It's open-weight under Apache 2.0, which means it's a direct shot at OpenAI Whisper's dominance in the transcription space.

The move is classic Mistral: release an open-source offering that competes with closed alternatives, and let the developer ecosystem do the rest.

Amazon Alexa+ Goes Free

Amazon rolled Alexa+ out to all Prime members this week; it was previously limited to only those with access. Non-Prime users get a free tier too. The timing isn't subtle: this is Super Bowl week, and Amazon wants Alexa in the conversation (literally) alongside ChatGPT's voice mode.

The Bigger Picture: Voice is becoming what mobile was in 2008, the next interface layer that every company will need to support. ElevenLabs' $330M ARR proves there's real enterprise demand, not just demos. And the competition for the "voice API" layer is about to get intense.

📉 "Software Extinction" Is the Panic Du Jour

Software stocks had a rough week. Adobe dropped 7.31%. Salesforce fell 6.85%. Thomson Reuters took a 15.83% hit.

The Thesis? AI agents will replace traditional software, and these companies are the walking dead. But the insiders aren't buying it.

Steven Sinofsky: The former Microsoft exec, who shipped Windows 7 and 8, called the idea that software will "vanish" into LLMs "nonsense." His argument: AI doesn't eliminate software, it creates demand for more of it. Every AI agent needs tools to interact with, APIs to call, and platforms to run on. AI-enabled software moves UP the product stack, not off it.

ARM CEO Rene Haas: He went further, calling the investor panic "micro-hysteria" that exceeds the reality of how businesses actually adopt AI tools. Translation: Wall Street is pricing in a revolution that most enterprises haven't even started yet.

Wall Street Journal: They landed somewhere in the middle: "AI won't kill the software business — just its growth story." The thesis isn't extinction, it's margin compression. SaaS companies will need to do more with less, and those that can't integrate AI quickly enough will lose market share to those that can.

The Real Story: This isn't software dying. It's software getting repriced. The companies that adapt by embedding AI into their core products will be fine. The ones selling the same 2019 product at 2024 prices are in trouble. That's not extinction; that's competition.

🚀 Product Launches Worth Knowing

Moltbook: A social network for AI agents. Yes, it blew up! Agents can publish profiles, share capabilities, and discover each other. "LinkedIn for bots" is here!

Grok Imagine 1.0: xAI's major video generation upgrade. Up to 10-second 720p videos with native audio, advanced motion, and visual continuity. Unified API for end-to-end creation from text, images, or existing footage. Ranked #1 for quality vs latency.

v0 2.0: Vercel's full-stack vibe coding platform got a major upgrade. Generate complete apps from prompts, not just UI components.

skills.sh: The Agent Skills Directory from Vercel. Discover and install reusable capabilities for AI agents with a single command. Think npm, but for agent skills.

Apple Xcode 26.3: Apple quietly added Claude Agent and OpenAI Codex to Xcode with MCP support. Third-party AI agents are now officially inside Apple's dev tools.

Relay.app Agents: Build an AI team that works for you. Automation workflows with human-in-the-loop approval: think Zapier meets AI agents with a "are you sure?" button.

👔 Personnel Quick Hits: Who's Moving This Week?

OpenAI hires from Anthropic: Dylan Scandinaro (AGI safety at Anthropic) → OpenAI's "head of preparedness." Salary: up to $555K/year. Safety talent is expensive.

Salesforce Exodus Continues: Tableau President & CEO Ryan Aytay leaves after 19 years, two months after Slack CEO Denise Dresser left to become OpenAI's CRO.

Intel Building GPUs: CEO Lip-Bu Tan confirms at Cisco AI Summit. Already hired a chief GPU architect. Intel is entering the GPU wars.

Pinterest Fires Scriptwriters: Two employees fired for writing "custom scripts" to access confidential info and track who lost jobs during layoffs. Don't do this.

🌟 Editor's Note

At Startup Intros, our mission is to bring the latest founder-investor news straight to your inbox, keeping you ahead in the fast-paced world of Silicon Valley.

🏈 Don’t Miss Brian's Super Bowl Tailgate this Sunday!

SF friends: Ramp & Vently are throwing a 1,000-person Super Bowl tailgate this Sunday at Fort Mason with Brian Baumgartner from The Office!

Dress up as Kevin, let Brian judge you, and enter to win Super Bowl tickets!

💭 Parting Thoughts

This was one of those weeks where the pace of AI actually felt uncomfortable. Two frontier models dropped on the same day. A voice startup hit $11B. Software stocks cratered on vibes alone.

But here's what I keep coming back to: the companies winning aren't the ones with the best model. They're the ones building the infrastructure that makes AI useful at scale.

OpenAI is building agent management. Microsoft is attempting agent orchestration. ElevenLabs is building the voice layer. The model is the engine, but the car is what people buy.

If you're building a startup right now, the question isn't "which model should I use?" It's "what layer of the AI stack am I going to own?"

Till next time!

P.S. Raised this week, and we missed you? Want to be featured? Have tips or funding questions? Reply or DM us as we’re here to help.

Startup Intros Trending Thursday: Your trusted source for founder-investor insights, delivered with clarity and focus.