Trending Year in Review #2025

From Vibe Coding to Code Red: The AI moments that defined 2025

Hi !

Before we dive into the best AI moments of 2025, I've partnered with Notion to drop $50K in founder tools, including credits from PostHog, v0, and Supabase, with a vetted list of investors to help you go from idea to funded startup.

Here's what's been trending in AI & tech from this past year:

  • DeepSeek R1 wiped $600B off Nvidia in a single day, and forced Silicon Valley to question everything about scaling

  • Vibe Coding went from Karpathy tweet to Collins Dictionary Word of the Year, and spawned a $50B+ industry

  • Claude Code hit $1B ARR in 6 months; Cursor crossed $1B ARR and $29.3B valuation; and Lovable hit $6.6B

  • Google declared "Code Red" in 2022 over ChatGPT. In 2025, the roles reversed as Sam Altman issued his own

  • Ghibli filter broke the internet, Nano Banana conquered image editing, and AI became culture

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.

🎁 Megadrop: $50K in Founder AI Tools

I’m excited to partner with Notion to give you $50K in AI Founder tools to help startup founders go from idea to MVP to raising funds with confidence!

What you get:

  • $50k in PostHog credit to track real user behavior

  • $60 in v0 by Vercel credit to vibe code your front-end

  • $50 in Supabase credit to spin up your backend

  • Up to 6 months of Notion Biz + AI to run your startup

But wait, there's more:

  • A Vetted List of 2,000+ early-stage VCs & accelerators

  • 50 Pitch Decks from YC, a16z & Sequoia startups

  • Guides on pitching and raising from the best founders!

  • Comment "Founder" on my LinkedIn post, connect with me & I'll DM you the tools!

Why Notion? They just crossed $600M in ARR (half from AI products) and are preparing for an IPO at an $11B valuation. They know what founders need to scale.

Repost & Share it & I'll jump on a 15-min call with you (for the first 40 reposts)!

πŸ”₯ Context VC & Startup Intros Events Coming Up

πŸ’₯ DeepSeek's $600B Wakeup Call

The year started with a shock.

On January 20, a Chinese AI lab called DeepSeek released R1, an open-weight reasoning model that rivaled OpenAI's o1:

  • The Claimed Training Cost: Less than $6M (versus OpenAI's $100M+ for GPT-4).

  • The Hardware: Just 2,048 Nvidia H800 chips (mid-range, export-restricted processors).

  • The Result: The largest single-day market cap loss in US history.

Nvidia lost $589-600B in one day, or a 17% decline. The Nasdaq lost $T. Broadcom fell 17%.

Marc Andreessen called it "AI's Sputnik moment."

The implications were existential for Silicon Valley's entire thesis

  • What if you don't need $100B data centers? 

  • What if compute efficiency matters more than raw scale?

  • What if the "moat" everyone believed in, massive infrastructure spending, was actually a liability?

DeepSeek became the #1 free app on the App Store, briefly beating ChatGPT. The company demonstrated that smart training techniques (mixture-of-experts, reinforcement learning) could achieve frontier performance at a fraction of the cost.

The stock recovered. The question didn't: Is scaling all you need?

πŸ’š Nvidia's $4T Coronation

If DeepSeek was the year's biggest scare, Nvidia's trajectory was its biggest flex.

In July, the company became the first in history to hit $4T in market cap. By October, it briefly touched $5T. The trajectory:

  • $1T in May 2023

  • $2T in February 2024

  • $3T in June 2024

  • $4T in July 2025

That's adding $3T in just over two years, which is more than the entire UK stock market is worth.

But the year wasn't easy. US export restrictions on China created massive headwinds:

  • Trump blocked the H20 chip (explicitly designed to comply with existing restrictions)

  • Nvidia took $4.5B+ in write-downs related to China exports

  • CEO Jensen Huang said the "$50B China market is effectively closed to US industry".

  • Chinese companies found workarounds (Tencent acquired Blackwell access through a Japanese data center)

The restrictions forced uncomfortable negotiations. In August, chip designers agreed to share 15% of Chinese sales with the US government. In December, Trump authorized H200 sales to "approved customers" with a 25% revenue share.

Despite it all, Nvidia's revenue soared to $187.1B in 2025. The AI infra buildout was simply too massive to slow down.

πŸ§‘β€πŸ’» Vibe Coding: From Tweet to Software 3.0

In February, Andrej Karpathy tweeted what became the defining tweet of the year: describing a new programming paradigm where you "fully give in to the vibes, embrace exponentials, and forget that the code even exists."

The post got 4.5M+ views. Within weeks, Wikipedia had a page. By March, Merriam-Webster added it as "slang & trending." In November, Collins Dictionary crowned it Word of the Year 2025.

But Karpathy wasn't done. At YC’s AI Startup School in June, he expanded the idea into Software 3.0:

  • Software 1.0: Traditional code (Python, C++)

  • Software 2.0: Neural networks with learned weights

  • Software 3.0: LLMs programmed in natural language

"Prompts are now programs that program the LLM," Karpathy explained. "And remarkably, these prompts are written in English."

The numbers backed it up: YC reported that 25% of startups in their Winter 2025 batch had codebases that were 95% AI-generated. Combined valuations of vibe coding startups jumped from ~$8B to over $50B in a single year.

Vibe coding isn't just a meme anymore; it's the name for how millions of people now build software.

πŸ’» The Developer Tools Explosion

Vibe coding created a gold rush. And the biggest winners were the VS Code forks or front-end builders.

Cursor is the category leader:

But Cursor's not alone:

The combined valuations now exceed $50B. Each of these companies is racing toward the same vision: a world where describing software is as easy as building it.

πŸš€ Claude Code: Fastest to $1B Ever

Among all the developer tools, one stood out.

In May, Anthropic launched Claude Code. By November, it had crossed $1B in ARR, the fastest dev tool to hit that milestone in history. The numbers:

  • $1B ARR in 6 months (represents ~10% of Anthropic's total run rate)

  • Usage grew 10x in 3 months

  • Enterprise clients include Netflix, Spotify, KPMG, L'Oreal, and Salesforce

  • Karpathy called it "the first convincing demonstration of what an LLM Agent looks like."

What separates Claude Code from competitors is that it runs locally in a private environment, gives developers terminal-level control, and handles complex multi-file edits autonomously. It's not an autocomplete tool; it's an agent that can execute, test, and iterate.

On December 3, Anthropic made its first major acquisition, Bun, the JavaScript runtime with 7M monthly downloads.

The message was clear: Claude Code isn't just a product, it's a platform bet.

πŸ”„ Google & ChatGPT: Trading Places

Three years ago, Google sounded "Code Red" over ChatGPT. CEO Sundar Pichai warned it could threaten the future of Search.

In December 2025, the roles reversed completely.

Sam Altman issued his own "Code Red" memo, marshaling resources to improve ChatGPT as competitive pressure from Gemini 3 intensified.

Google's 2025:

  • Spent the early year recovering from embarrassing Gemini 1.0 failures

  • Released Gemini 3 in November with strong benchmark results

  • Grew to 650M monthly Gemini users

  • Deployed across Search, Workspace, and Android instantly

  • Salesforce CEO Marc Benioff publicly said he was ditching ChatGPT for Gemini

OpenAI's 2025:

  • GPT-5 launched in August to "mixed reviews"

  • Dozens of top researchers left for Mira Murati's Thinking Machines and Meta's Superintelligence Labs

  • Lost enterprise market share: fell to 27% vs. Anthropic's 40% and Gemini's 21%

  • Still dominates consumer: 800M weekly active users

Altman's Code Red memo delayed ads, shopping agents, and personalization features, all to focus on ChatGPT's core experience. OpenAI released GPT-5.2 in December, just weeks after the memo.

The irony is Thick: Google invented the transformer architecture. For years, Google was the undisputed leader in AI research. Then ChatGPT happened, and Google spent three years playing catch-up.

Now, finally, they're making OpenAI sweat.

🎭 The Meme Economy: When AI Went Viral

Beyond the valuations and policy fights, 2025 was the year AI became culture.

The Ghibli Filter (March):

ChatGPT's GPT-4o image generation went viral when users discovered it could transform any photo into Studio Ghibli-style art. OpenAI added 1M users per hour during the peak.

Sam Altman tweeted: "Can yall please chill on generating images this is insane our team needs sleep."

Everything got Ghibli-fied: historic photos, political moments, classic memes. Hayao Miyazaki's 2016 quote resurfaced: AI art is "an insult to life itself." Copyright lawyers took notes.

Nano Banana (August):

An anonymous model called "nano-banana" appeared on LMArena and dominated image editing benchmarks. Two weeks later, Google confirmed it was Gemini 2.5 Flash Image, and the figurine trend exploded. Turn your selfie into a hyperrealistic 3D toy.

  • 13M new Gemini users in 4 days

  • 200M image edits within weeks

  • November: Nano Banana Pro launched with infographics and multi-image fusion

Google's stealth launch strategy created organic buzz in a way their marketing never could.

2025 was the year AI stopped being a technology story and became a cultural one. The memes, the filters, the figurines, these are the things people will remember.

πŸ“° Quick News: Nvidia's $20B Groq Deal + Trump's Tech Trade War

Nvidia Eyes $20B Groq Acquisition: The AI infrastructure giant is in advanced talks to acquire Groq, the LPU chip startup known for ultra-fast inference speeds. The $20B deal would give Nvidia a specialized inference chip to complement its training-focused GPUs, and eliminate a potential competitor before it scales. Groq's Language Processing Units can run LLaMA 2 at 500+ tokens/second, making it popular for real-time applications.

Trump's Tech Trade War Escalates: The administration is preparing new export controls that would restrict AI chip sales beyond China to include additional countries deemed security risks. Nvidia disclosed that existing China restrictions already cost $8B in lost H20 sales. The 25% revenue-sharing arrangement for approved H200 sales signals a new era of "managed access" to U.S. AI technology.

🌟 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.

🎁 $50K Founder Tools Giveaway

Partnering with Notion (just crossed $600M ARR, prepping for IPO) to give you: PostHog, Vercel & Supabase credits, 6 months of Notion Biz + AI, 2,000+ VC contacts, and 50 pitch decks from YC/a16z startups.

Comment "Founder" on my LinkedIn post to grab it.

πŸ’­ Parting Thoughts

Every year has defining moments. 2025 had defining forces.

DeepSeek proved efficiency can beat scale. Nvidia proved infrastructure always wins anyway. Vibe coding proved that anyone can build software now. Cursor and Lovable proved that non-technical founders can ship. Claude Code proved that agents actually work. Gemini 3 proved Google still fights. The Ghibli filter proved that AI is culture now.

The through-line: AI stopped being something we talk about and became something we talk through. The tools, the policies, the memes; they're all evidence of the same shift.

We're not covering AI anymore. We're living in it.

Till next time!

Dev Chandra
Founder & CEO @ Startup Intros
Associate @ Context VC
LinkedIn: /in/devchandra

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.