I've been the biggest Google hater for years, it was completely mismanaged, destroyed by politics and lack of any leadership, fumbled inventing Transformers etc.
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS69Pj AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPegSl
How would you like to sponsor a novel I would like to write? The novel shows the unfolding reality of a one person billion-dollar company. Fictional scenario. 50K.
We’re excited to partner with @intelligenceco. They’re building Cofounder, an autonomous agent that can integrate with every system within a startup (Notion, Slack, Gmail, Github), ingest global company context, and take actions on behalf… https://t.co/igm1W8RRmupic.twitter.com/Dw7hrKFInP
I met johnny a few weeks ago after @vanlancker emailed me with his github and links to some of his tweets.
I was immediately impressed. Omni was beautifully made, with a clear product direction and sophisticated agent work done. The next day (the day before thanksgiving) we… https://t.co/IiKUgugDUr
when I joined @Lovable after leaving @stripe we had 4 active subscribers
today we announce our $330M series b but it still feels like day 0. tons of very difficult problems that require brilliant engineers. unprecedented opportunity to have impact. join me…
Since becoming active on X when I started investing (~2021), this platform has helped me: • Meet founders I later partnered with • Make a few of my closest friends • Connect incredible engineers & operators to the companies I work with
The Next Frontier in AI: Why World Models Could Rival LLMs in Importance
On December 18, 2025, Vinod Khosla—one of Silicon Valley’s most influential venture capitalists—posted a deceptively simple sentence on X:
“World models will be as important as LLMs. Next big market and General Intuition has the best data set for this.”
Coming from the man who backed OpenAI early and has consistently anticipated major platform shifts, the statement landed like a flare shot into the AI night sky. It signaled something profound: the age of language-first artificial intelligence may be giving way to something deeper, more physical, and far more consequential.
If large language models (LLMs) taught machines how to talk, world models aim to teach them how to understand reality itself.
The Age of LLMs: A Brief Reckoning
Since roughly 2022, LLMs have dominated the AI conversation. Models like GPT-4, Claude, and Grok transformed how humans interact with machines—turning natural language into a universal interface for software, creativity, and reasoning.
LLMs became:
Customer support agents
Code-writing copilots
Tutors, translators, marketers, and analysts
They excelled because language is compressed human knowledge. Train a model on enough text, and it begins to reason, infer, and generalize. The result was an explosion of productivity—and a multi-billion-dollar industry almost overnight.
But LLMs have a fundamental limitation: they live in symbols, not in space and time.
An LLM can describe how to drive a car, assemble a robot, or perform surgery—but it cannot simulate the act. It lacks intuition about gravity, momentum, friction, or causality. It knows the words for the world, not the world itself.
This is the ceiling world models aim to break through.
What Are World Models, Really?
At their core, world models are AI systems that learn how environments behave over time.
Instead of predicting the next word in a sentence, they predict the next state of the world.
They ingest:
Video
Images
Sensor data
Action–reaction sequences
And they learn the underlying dynamics—how objects move, collide, deform, disappear, or transform when actions are taken.
If LLMs are encyclopedias that talk, world models are internal physics engines.
Think of them as a simulator running inside an AI’s mind:
If I do this, what happens next?
If that object falls, where does it land?
If the environment changes, how should I adapt?
This idea draws from reinforcement learning, cognitive science, and neuroscience. Humans constantly run world models subconsciously—we imagine outcomes before acting. World models attempt to give machines the same ability.
Why World Models Matter More Than You Think
World models unlock embodied intelligence—AI that doesn’t just respond, but acts.
This has sweeping implications:
1. Robotics and Manufacturing
Robots trained with world models can rehearse millions of scenarios before ever touching a factory floor. Assembly lines become simulations first, reality second.
2. Autonomous Vehicles
Instead of reacting frame by frame, cars can simulate traffic futures—anticipating accidents before they occur.
3. Healthcare and Drug Discovery
World models can simulate molecular interactions, disease progression, or treatment outcomes—compressing years of lab work into days.
4. Science and Climate
From weather systems to particle physics, world models allow researchers to explore “what-if” universes at scale.
5. Gaming and Entertainment
Infinite, adaptive worlds where NPCs behave less like scripts and more like sentient actors.
In short, LLMs made AI conversational; world models make it situational.
From Words to Worlds: The Shift to Multimodal Intelligence
We’ve already seen early signs of this transition.
Google’s Dreamer showed how world models enable agents to learn efficiently with limited data.
OpenAI’s Sora demonstrated video generation that respects physical consistency—objects persist, gravity holds, time flows.
Diffusion-based and transformer-based world models now allow AI to generalize from sparse experiences.
These systems don’t memorize scenes—they infer laws.
That’s the leap from parroting reality to understanding it.
General Intuition: Why Khosla Is Betting Big
Khosla’s tweet wasn’t abstract. It pointed directly to General Intuition, a startup he describes as his biggest bet since LLMs.
Founded in 2025 as a public-benefit corporation and based in New York and Geneva, General Intuition raised $134 million in seed funding—one of the largest seed rounds in AI history.
What makes it special isn’t just talent or architecture.
It’s data.
The Medal Data Moat
General Intuition has access to trillions of video interactions from Medal, the world’s largest gaming clip platform.
Games are not trivial environments. They are:
High-dimensional
Physics-rich
Adversarial
Temporally complex
In other words, they are perfect training grounds for intelligence.
A game demands:
Spatial reasoning
Anticipation
Strategy
Improvisation under uncertainty
Exactly the skills required in the real world.
As one researcher put it: games are flight simulators for intelligence itself.
Inside General Intuition’s Stack
The company is building across three layers:
World Models: Systems like GAIA-2 for video generation and ฮ-IRIS for efficient environment tokenization.
Agents: AI entities capable of planning, adapting, and reasoning within simulated worlds.
Video Understanding: Extracting transferable knowledge from gameplay into real-world domains.
Early demos show agents mastering generated shooter games at professional levels—learning purely from interaction, not instruction.
By late 2025, the company was reportedly raising additional capital at a valuation exceeding $2 billion.
To Khosla, this isn’t a niche play. It’s infrastructure.
World Models vs. LLMs: Rivalry or Convergence?
The real insight isn’t that world models will replace LLMs.
It’s that the future belongs to systems that combine both.
LLMs provide abstraction, planning, and language.
World models provide intuition, simulation, and grounding.
Together, they form something closer to general intelligence.
Language becomes the interface. World models become the reality engine.
Risks, Challenges, and Ethical Questions
This future is not without friction.
Compute costs remain immense.
Data ownership becomes thorny when training on user-generated content.
Simulation bias can lead to flawed real-world decisions.
Ethical concerns arise when simulated outcomes influence medical, legal, or military actions.
As simulations grow more convincing, the line between prediction and prescription blurs.
World models must be governed as carefully as they are built.
The Bigger Picture: Why This Shift Is Inevitable
LLMs taught machines to speak our language.
World models will teach them to live in our world.
If the last AI revolution turned text into intelligence, the next one will turn experience into intuition.
Vinod Khosla’s message is ultimately a warning and an invitation: The center of gravity in AI is moving—from words to worlds.
Those who understand that shift early won’t just build better models. They’ll build the operating system for reality itself.
And that may prove even more transformative than language.
From Trial and Error to Imagination: Reinforcement Learning, World Models, and the Future of Intelligent AI
Artificial intelligence did not begin with language models. Long before machines learned to write essays or generate code, researchers were trying to answer a deeper question:
How can a machine learn to act intelligently in the world?
The answer to that question gave rise to Reinforcement Learning (RL)—and, more recently, to its powerful evolution: reinforcement learning powered by world models. Together, they are reshaping how AI learns, plans, and ultimately behaves in complex, real-world environments.
What Is Reinforcement Learning?
Reinforcement Learning is a branch of machine learning focused on decision-making over time. Instead of learning from labeled examples, an RL system learns by interacting with an environment, observing the consequences of its actions, and improving through feedback.
At its core, RL is learning by experience—much like how animals or humans learn to walk, drive, or play a game.
The Core Components of Reinforcement Learning
Every RL system is built around a simple but powerful loop:
Agent The learner or decision-maker (for example, a robot, a game-playing AI, or a trading algorithm).
Environment The external world the agent interacts with, which responds to actions.
State (s) A snapshot of the current situation (such as a game board position or sensor readings).
Action (a) A choice the agent can make (move left, accelerate, buy a stock).
Reward (r) Immediate feedback from the environment—positive for success, negative for failure.
Policy (ฯ) The agent’s strategy: a mapping from states to actions.
Value Function (V or Q) An estimate of how good a state or action is in terms of long-term reward.
The agent starts out naรฏve, often acting randomly. Over time, it refines its behavior using algorithms such as Q-learning, policy gradients, or actor–critic methods, balancing:
Exploration (trying new actions)
Exploitation (using what already works)
Classic successes of RL include AlphaGo, robotic locomotion, and game-playing agents—but these systems came with a major limitation.
The Fundamental Limitation of Classical Reinforcement Learning
Traditional RL is expensive, slow, and often unsafe.
Robots wear out when trained by trial and error
Self-driving cars cannot “experiment” freely in traffic
Real-world environments offer limited opportunities for exploration
Learning from raw sensory data (like video) is extremely sample-inefficient
In short, brute-force experience doesn’t scale.
To move forward, AI needed something humans already have:
Imagination.
What Are World Models?
World models are AI systems that learn how an environment works internally—not just what actions yield rewards, but how the world evolves over time.
If reinforcement learning answers “What should I do?”, world models answer “What will happen if I do it?”
The Intuition Behind World Models
A world model is like a mental simulator inside the agent’s mind.
Humans constantly run such simulations:
If I step here, will I slip?
If I turn now, will I miss the exit?
If I say this, how might they react?
World models give AI the same ability.
How World Models Work
Technically, world models:
Compress high-dimensional observations (images, video, sensor data) into a latent representation
Learn the environment’s dynamics:
Predict future states, rewards, and sometimes uncertainty
Generate hypothetical futures without real-world interaction
They are often built using:
Variational autoencoders (VAEs)
Transformers
Recurrent neural networks
Diffusion-based video models
Modern examples include video-prediction systems like Sora, learned physics simulators, and environment models trained from gameplay or sensor streams.
In essence, world models turn experience into foresight.
Reinforcement Learning Inside World Models: The Big Leap
When reinforcement learning is combined with world models, the result is model-based reinforcement learning—a paradigm shift in how AI learns.
Instead of learning only from real-world experience, the agent learns by practicing inside its own imagination.
Step-by-Step: How RL Works with World Models
1. Learning the World Model
The agent first collects real interaction data:
State
Action
Next state
Reward
A neural network is trained to predict what comes next. Over time, it learns the rules of the environment.
This is decision-making by foresight rather than reflex.
3. Training From Imaginary Experience
The agent generates synthetic data inside the model and uses it to update its policy and value functions—dramatically reducing the need for real-world trials.
4. Reality Check and Iteration
The agent alternates between:
Real-world data collection
Model refinement
Simulated training
To avoid being misled by model errors, techniques like uncertainty estimation and ensemble models are used.
Why World-Model-Based RL Is So Powerful
1. Sample Efficiency
A robot can “practice” millions of times without physical wear and tear.
2. Safety
Dangerous or costly actions are tested in simulation first.
3. Generalization
World models allow agents to imagine variations and adapt to new situations.
4. Scalability
High-dimensional inputs like raw pixels become manageable through learned representations.
5. Long-Horizon Planning
Agents can plan far into the future—essential for navigation, strategy, and real-world autonomy.
6. Emergent Intelligence
Agents begin to exhibit behaviors that look like reasoning, anticipation, and creativity.
The downside? World models are computationally expensive, and inaccurate models can introduce bias. A flawed imagination leads to flawed decisions.
Real-World Systems That Use These Ideas
Dreamer (Google DeepMind)
An agent learns a world model from image sequences and trains policies almost entirely in imagination—achieving top-tier performance with minimal real interaction.
MuZero (DeepMind)
Learns game rules from scratch and plans using an internal model, mastering Go, chess, and Atari without explicit knowledge of the environment.
Robotics and Autonomous Driving
Companies like Tesla and OpenAI use learned simulations to train navigation and manipulation policies.
Gaming and Synthetic Worlds
Startups such as General Intuition use massive gameplay datasets to train world models that enable agents to master complex, dynamic environments.
Beyond Games
Drug discovery (molecular simulation)
Climate modeling
Finance and trading agents
Supply chain optimization
Anywhere decisions unfold over time, world-model-based RL applies.
From Brute Force to Intelligence
Traditional reinforcement learning is like learning by touching a hot stove repeatedly.
Reinforcement learning with world models is like learning by thinking first.
This shift—from reaction to prediction, from trial to imagination—is one of the most important transitions in modern AI.
It is how machines move from:
Acting → Planning
Reacting → Anticipating
Optimizing → Understanding
As AI systems increasingly operate in the real world, imagination will matter as much as experience.
World models don’t just make reinforcement learning faster. They make it smarter.
And that may be the key to truly autonomous intelligence.
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS69Pj AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPegSl
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS69Pj AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPegSl
I’m done doomscrolling. Give me lovebait instead. Bait me with kindness. Hook me with hope. Algorithm, show me strangers choosing grace, small wins, soft moments, and proof that people are still trying. ๐
I work with a lot of this generations high potential CEOs. One thing I've noticed about them is how HUMBLE they all are...literally all of them. I like this generation.
I want you and your company to sponsor work on a novel I will write. When it is finally time, there is a major debate inside SpaceX: Mars or Titan. 50K.
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS69Pj AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPegSl
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS6HER AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPeOHT
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS69Pj AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPegSl
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS6HER AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPeOHT
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS6HER AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPeOHT
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS69Pj AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPegSl
I'd like to do marketing for you. I have the equivalent of an Oscar in digital marketing. ๐ฃ️ Marketing: The New Bottleneck in the Age of AI https://t.co/b3TZRS6HER AI Has Commoditized Building, And Marketing Is Now The Real Choke Point https://t.co/Pb6IHPeOHT
I'd like to do marketing for you. Digital Marketing Minimum https://t.co/hki7izgnzZ Marketing Principles Plus AI https://t.co/UKWWipfmwK The AI Marketing Revolution: How Artificial Intelligence is Transforming Content, Creativity, and Customer Engagement https://t.co/EGSKPBJdAc
Beehiiv: The Infrastructure Behind the Modern Newsletter Economy
In the early days of the internet, blogs were campfires—small, personal, flickering points of light in a vast digital wilderness. Social media later arrived like megacities, loud and algorithmically dense, where creators rented attention but never owned the land. Newsletters, however, have become something else entirely: private railroads between creators and audiences. And Beehiiv is laying the tracks.
Beehiiv is not just another email tool. It is infrastructure—purpose-built for the newsletter economy, optimized for growth, monetization, and ownership in a world increasingly hostile to creators who depend on platforms they do not control.
What Beehiiv Is—and What It Is Not
Beehiiv is an all-in-one newsletter platform designed for creators, publishers, brands, and media companies to build, grow, and monetize audiences through email, websites, and integrated growth networks.
Unlike general email marketing platforms such as Mailchimp or ConvertKit—which evolved to serve businesses running campaigns, funnels, and CRM-style automation—Beehiiv was designed from first principles around newsletters as a media product. Its philosophy is clear: newsletters are not marketing artifacts; they are businesses.
At its core, Beehiiv offers:
A polished, intuitive newsletter editor
A no-code website builder optimized for SEO
Unlimited email sends (a crucial differentiator)
Advanced audience analytics and segmentation
Native referral programs and recommendation networks
Built-in monetization via an ad network and paid subscriptions
Surveys, polls, and engagement tools
Custom domains and best-in-class deliverability
AI-powered writing and image tools
Email automations, webhooks, and integrations
Team collaboration and enterprise-grade controls
It integrates seamlessly with Stripe, Zapier, Google Analytics, and leading CRMs—allowing creators to run sophisticated media operations without duct-taping together half a dozen tools.
Think of Beehiiv less as “Substack with features” and more as Shopify for newsletters—a platform that assumes ambition.
Pricing and Scale
Beehiiv’s pricing reflects its growth-first ethos:
Launch (Free): Up to 2,500 subscribers, unlimited sends, custom website
Scale: $43/month
Max: $96/month
Enterprise: Custom pricing for 100,000+ subscribers
As of Q3 2025, Beehiiv’s footprint is striking:
130,000+ publishers
46+ billion emails sent
425 million unique readers
$45+ million earned by creators
5/5 rating from 19,000+ users
Used by high-profile creators such as Arnold Schwarzenegger, JJ Redick, and Colin & Samir
The company operates 100% remotely, employing 109 full-time team members across 12 countries, speaking six languages—a global company serving a global creator class.
Founders: Built in the Furnace of Morning Brew
Beehiiv’s credibility comes from its lineage.
The company was founded by Tyler Denk (CEO) alongside Benjamin Hargett and Jacob Hurd, all early architects of Morning Brew, one of the most successful modern newsletters ever built.
Tyler Denk joined Morning Brew in 2017 as its second employee, wearing every hat imaginable—product, engineering, growth. He helped design the referral engine, SEO-optimized publishing stack, and analytics systems that scaled Morning Brew from 100,000 to over 3 million subscribers before its 2020 acquisition by Business Insider.
Benjamin Hargett was Morning Brew’s first full-time engineer, while Jacob Hurd joined as its fourth engineer, becoming deeply specialized in newsletter infrastructure at scale.
After the acquisition, Tyler briefly joined Google as a product lead on YouTube Music, but the gravitational pull of the creator economy proved stronger.
The Origin Story: Building in the Shadows
Beehiiv was born from frustration.
At Morning Brew, the founders had custom-built tools that gave them unfair advantages. Everywhere else, creators were stuck stitching together WordPress, Mailchimp, Airtable, Stripe, and prayer. The result: stalled growth and burnout.
In late 2020, the three founders reconnected over Google Meet and began building Beehiiv nights and weekends, quietly, obsessively. For nearly ten months, they bootstrapped an MVP in secret—sometimes spending random Saturdays in Brooklyn hacking on version one.
Their goal was radical in its simplicity: Democratize elite newsletter infrastructure. No algorithms. No gatekeepers. No dependency on Big Tech’s mood swings.
Tragedy and Resilience
In 2021, Andrew Platkin—an early Morning Brew mentor—joined Beehiiv as a technical advisor and later became CTO after the seed round. He was instrumental in shaping Beehiiv’s early architecture.
On April 29, 2022, Andrew passed away suddenly after feeling unwell.
It was a devastating loss—emotionally and operationally. Features were delayed. Systems strained. The team faltered—but did not fracture. His equity was accelerated, and his mother became a shareholder. Beehiiv carried forward with a renewed sense of purpose.
Funding in a Skeptical Market
Despite growing competition—Twitter acquiring Revue, Facebook launching Bulletin—Tyler raised a $2.6 million seed round in July 2021 in just weeks. Investors included:
Social Leverage
Thirty Five Ventures (Kevin Durant’s firm)
Company Ventures
Adventure Fund
Spice Capital
Bullish Studio
Beehiiv officially launched on November 17, 2021, announced via a single tweet. The product resonated immediately.
Growth, Profitability, and the Ad Network Flywheel
By mid-2025, Beehiiv had reached $30+ million in ARR, up from $19.8M at the end of 2024, serving over 90,000 paying customers.
Revenue composition is telling:
~$20M from software subscriptions
The remainder from ads and monetization tools
The platform now processes 800 million monthly impressions and is profitable—a rarity in the creator-economy SaaS space. Its valuation stands around $225 million.
One key differentiator: Beehiiv takes 0% on paid subscriptions on higher tiers, aligning incentives with creators rather than taxing success.
Turning Points That Defined Beehiiv
Several moments reshaped Beehiiv’s trajectory:
Late 2020: Ideation and bootstrapping begin
July 2021: $2.6M seed round closes
November 2021: Public launch
April 2022: Loss of CTO Andrew Platkin
September 2022: $1.6M seed extension
May 2023: $12.5M Series A led by Lightspeed (raised in under a week)
Q3 2023: Acquisition of Swapstack; AI tools launched; full rebrand
2024–2025: Series B ($33M led by NEA); $30M ARR milestone; 130K+ publishers
Notably, Beehiiv hit profitability in April 2023, choosing sustainability over growth-at-all-costs—a contrarian move in venture-backed SaaS.
Why Beehiiv Matters
Beehiiv’s rise mirrors a broader shift.
Creators are leaving algorithmic feudalism—where platforms dictate reach—and moving toward owned audiences. Email, once dismissed as legacy tech, has become the most reliable, high-signal channel in digital media.
Beehiiv doesn’t just ride this wave—it shapes it.
It treats newsletters not as side projects, but as media companies in miniature, deserving of professional tools, revenue streams, and dignity.
In an internet defined by noise, Beehiiv is building something quieter, stronger, and far more durable: direct relationships at scale.
And in the long run, infrastructure always wins.
Beehiiv’s AI Toolkit: Power Tools for the Modern Newsletter Creator
In the creator economy, attention is scarce—but time is scarcer. Every newsletter writer knows the feeling: staring at a blank editor, juggling ideas, formatting visuals, polishing tone, and wondering whether today’s issue is good enough to hit “send.” Beehiiv’s AI tools are designed for exactly that moment—not to replace the creator, but to remove friction from the creative process.
Rather than bolting generic AI onto an email platform, Beehiiv has embedded artificial intelligence directly into the newsletter workflow itself, turning AI into a silent co-pilot—always present, never intrusive.
AI Built Inside the Newsletter, Not On Top of It
Beehiiv’s AI tools live directly within the post editor, not behind a separate dashboard or third-party integration. This design choice matters. It means creators can write, edit, visualize, translate, and publish without context-switching—a critical advantage in maintaining creative momentum.
Launched as part of Beehiiv’s broader product ecosystem, the AI suite is intentionally newsletter-first, unlike generic AI writing tools that assume blogs, essays, or marketing copy. Every feature is optimized for email readability, cadence, and audience engagement.
The tools are available on Scale and Enterprise plans, with daily AI request limits that reset every 24 hours. These limits are shared across team members and publications—encouraging thoughtful use rather than mindless generation.
Importantly, Beehiiv’s AI emphasizes human control. Every output is fully editable. Nothing is locked. Nothing is auto-published. The creator remains the editor-in-chief.
What Beehiiv’s AI Actually Does
At a high level, Beehiiv’s AI toolkit focuses on four core pillars:
Text generation
Text editing and refinement
Image creation
Translation and localization
Recent expansions in November 2025 extended these capabilities into AI-powered website building and creator commerce, signaling Beehiiv’s ambition to become a full-stack creator platform rather than “just” a newsletter tool.
AI Writing Assistant: Beating the Blank Page
The flagship feature is the AI Writing Assistant, designed to eliminate the most painful part of writing: starting.
What It Does
The AI Writing Assistant can generate:
Full newsletter drafts
Individual sections
Introductions, summaries, headlines, or CTAs
Users provide a prompt—such as “Write a weekly product update in a friendly, concise tone”—and the AI generates editable text in real time.
How It Works
Inside the editor, creators insert an AI block, enter a prompt, optionally select parameters like:
Tone (professional, casual, conversational, etc.)
Length (words, paragraphs, characters)
With one click, the text appears directly in the draft, ready for refinement.
Why It Matters
This isn’t about outsourcing thinking. It’s about momentum. The AI acts like a junior writer throwing ideas onto the page—giving creators something concrete to react to, sharpen, or discard.
For experienced writers, it’s a perspective generator. For new writers, it’s a confidence builder. For teams, it’s a speed multiplier.
AI Text Tools: Editing at the Speed of Thought
If the Writing Assistant helps you create, Beehiiv’s AI Text Tools help you polish.
These tools are applied to existing text—either written by the creator or generated by AI—and focus on refinement rather than invention.
Capabilities Include:
Auto-complete sentences to finish half-formed thoughts
Fix spelling and grammar (currently English-only)
Change length (shorten, extend, or simplify)
Alter tone (formal ↔ casual, concise ↔ expressive)
How It Works
Creators highlight any text, click the AI icon in the toolbar, and choose an action. The revised text appears inline after a brief processing moment, with options to further tweak before confirming.
The Bigger Picture
This turns editing into a non-destructive, exploratory process. You can test multiple versions of a paragraph without rewriting it manually—much like adjusting exposure or contrast in photo editing rather than reshooting the image.
The result: fewer mistakes, tighter prose, and newsletters that feel intentional rather than rushed.
AI Image Tools: Visuals Without the Design Bottleneck
Newsletters increasingly compete not just on words, but on visual identity. Beehiiv’s AI Image Generator allows creators to produce custom visuals without leaving the editor—or opening Photoshop.
What It Can Create
Header images
Section dividers
Thumbnails
Featured visuals
Creators describe the image and select a style, including:
Photorealistic
Digital art
Comic book
Neon punk
Isometric
Line art
3D model
How It Works
Typing / in the editor opens the menu. Selecting AI Image prompts the user to enter a description and choose a style. Images generate in seconds and can be inserted, discarded, or regenerated.
Why This Matters
Images are no longer a creative bottleneck. Creators can maintain visual consistency and originality without stock photo fatigue or external tools—keeping production fast and branding cohesive.
AI Translator: Global Reach, Zero Friction
Email is global by default—but language often isn’t. Beehiiv’s AI Translator lowers that barrier.
Supported Languages
Including:
English
Spanish
French
German
Italian
Portuguese
Brazilian Portuguese
How It Works
Creators highlight text, select “Translate,” choose a language, and the translated version replaces the original text—no exporting, copying, or re-uploading required.
While it doesn’t change the publication’s default language settings, it enables rapid localization of content for international audiences.
Strategic Impact
This opens doors to new subscriber markets with almost no additional operational cost—especially valuable for publishers testing global expansion.
November 2025: From Newsletter Platform to Creator OS
In November 2025, Beehiiv made a decisive leap by integrating AI-powered website building, following its acquisition of Typedream, an AI-native site builder.
AI Website Building
Creators can now:
Build full websites via conversational AI
Recreate sites by uploading screenshots
Use drag-and-drop tools with professional templates
All of this integrates directly with newsletters for unified branding and SEO.
Additional Creator Tools
Beehiiv also expanded into:
Selling digital products (0% Beehiiv commission)
Podcast hosting and distribution
Real-time cross-channel analytics
“Link in bio” pages
Advanced audience segmentation
Automated workflows (e.g., welcome emails)
Enhanced ad dashboards with payout estimates
Together, these features position Beehiiv as a serious competitor to WordPress, Patreon, and Gumroad—but with newsletters at the center rather than as an afterthought.
Using Beehiiv’s AI: A Simple Workflow
Getting started is intentionally frictionless:
Open a post draft (or type beehiiv.new in your browser)
Type / or /AI to access AI tools
Generate text, edit sections, or create images
Translate or refine as needed
Monitor your daily AI usage in the prompt box
Everything processes in seconds. Everything is editable. Nothing is permanent until you hit publish.
The Philosophy Beneath the Features
Beehiiv’s AI tools are not about automation for its own sake. They are about removing invisible tax—the small frictions that slow creators down, drain energy, and break flow.
In an era where AI often feels like a blunt instrument, Beehiiv treats it as a scalpel: precise, optional, and always guided by human judgment.
The result is not less creativity—but more room for it.
Beehiiv AI vs. Substack AI: Two Philosophies of Augmented Creativity
In the modern creator economy, AI has become less of a looming replacement and more of an exoskeleton—amplifying human effort rather than supplanting it. Nowhere is this clearer than in the divergent AI strategies of Beehiiv and Substack, two of the most influential newsletter platforms shaping how ideas travel across the internet.
Both platforms embrace AI. But they embrace it for very different reasons.
Beehiiv treats AI as infrastructure for writing, growth, and monetization—a set of power tools embedded directly into the act of publishing. Substack, by contrast, treats AI as a force multiplier for accessibility and multimedia distribution, especially in audio and video.
The result is not a question of “which AI is better,” but which philosophy aligns with how you create.
Two Platforms, Two AI Worldviews
Beehiiv’s AI suite is deeply integrated into the newsletter editor itself. Launched in 2023 and significantly expanded in November 2025 following Beehiiv’s acquisition of the AI website builder Typedream, it reflects an ambition to become a full-stack creator operating system. Writing, editing, visuals, translation, websites, analytics, monetization—AI touches all of it.
Substack’s AI evolution began earlier, around 2022–2023, but took a different path. Rather than automate writing, Substack focused on audio, video, and accessibility—turning text into sound, speech into text, and long-form content into short, shareable media.
Crucially, neither platform uses AI to replace creators. Both frame AI as “superpowers,” not substitutes.
As of December 2025:
Beehiiv AI is available on Scale ($43/month) and higher plans, with daily usage limits that reset every 24 hours.
Substack AI tools are free for all publishers, though some features roll out gradually across publications.
In both cases, creators retain full editorial control over AI outputs.
Writing vs. Repurposing: Where the AI Lives
The clearest difference between Beehiiv and Substack is where AI enters the workflow.
Beehiiv inserts AI at the point of creation. Substack inserts AI after creation, during distribution and repurposing.
That single distinction shapes everything else.
Beehiiv AI: A Writing Room With Power Tools
Beehiiv’s AI is designed for creators who live in text.
Text Generation and Editing
Beehiiv includes a robust AI Writing Assistant capable of generating:
Full newsletter drafts
Sections and outlines
Subject lines (reported to lift open rates by ~23%)
Summaries, intros, and CTAs
Creators report saving 2–3 hours per newsletter, effectively compressing the writing process by an order of magnitude.
Complementing this is a suite of AI text editing tools:
Auto-complete sentences
Fix grammar and spelling (English-only)
Shorten, extend, or simplify content
Shift tone (casual ↔ professional)
This makes editing feel less like surgery and more like sculpting—testing variations without starting over.
Substack, notably, offers no native AI for text generation or editing. Roughly 78% of Substack publishers who use AI rely on external tools like ChatGPT for ideation, proofreading, or rewriting.
Images, Visuals, and Design
Beehiiv
Beehiiv’s AI Image Tools generate custom visuals—headers, thumbnails, section dividers—directly from text prompts and predefined styles (photorealistic, digital art, comic book, neon punk, and more). The emphasis is on brand cohesion and speed, eliminating the need for external design tools.
Substack
Substack also offers an AI image generator, producing multiple image options per prompt directly in the editor. The tool is simpler and more stylistic—used by about 41% of AI-adopting Substack publishers—and primarily supports post visuals and branding accents rather than deep design workflows.
Audio Is Where Substack Pulls Ahead
If Beehiiv dominates the writing desk, Substack owns the recording studio.
Text-to-Speech (TTS)
Substack’s AI narration converts English posts into audio using multiple synthetic voices. Many publications have TTS enabled by default, allowing readers to listen via a play button in the app.
This dramatically improves:
Accessibility (especially for visually impaired readers)
Convenience (commuters, multitaskers)
Engagement time
Transcription and Video Tools
Substack’s AI also:
Generates clean, editable transcripts for podcasts and videos in about a minute
Syncs transcripts with timestamps and captions
Automatically creates short video clips (1–2 minutes) from long-form content
Produces audiograms—static videos with captions and audio—for social sharing
Creators report that AI-generated clips drive 2.5× faster audience growth on social platforms.
Beehiiv, by contrast, remains text-first. While it supports podcasts, it does not yet offer AI transcription, narration, or video clipping.
Translation, Websites, and the Creator Stack
Here, Beehiiv again pulls ahead in breadth.
AI Translation: Beehiiv can translate newsletter content into major languages (Spanish, French, German, Portuguese, and more), enabling rapid global expansion.
AI Website Builder: Following the Typedream acquisition, Beehiiv allows creators to build full websites via conversational AI, drag-and-drop tools, or even screenshots of existing sites.
Substack does not currently offer AI translation or website-building tools, focusing instead on accessibility features like captions and transcripts.
Strengths by Creator Type
Beehiiv AI Is Best For:
Text-heavy newsletters
Writers battling blank-page paralysis
Growth- and monetization-focused creators
Teams that value customization and speed
Creators building full media businesses (newsletters, sites, products)
Beehiiv’s AI is a productivity engine—designed to accelerate writing, polish output, and optimize performance.
Substack AI Is Best For:
Podcasters and video creators
Publishers repurposing long-form content
Accessibility-first publications
Community-driven creators prioritizing sharing
Substack’s AI excels at amplification, not authorship.
Limitations and Trade-Offs
No AI strategy is without friction.
Beehiiv’s limitations:
AI access is paywalled
Daily usage limits
Limited audio/video intelligence
Substack’s limitations:
No native writing AI
Heavy reliance on external tools
Some AI features (like TTS) are English-only or still rolling out
Image tools are relatively basic
Many creators ultimately adopt a hybrid workflow—using Beehiiv for writing and growth, Substack for community and multimedia, and external AI tools to fill the gaps.
Choosing the Right AI Philosophy
Beehiiv and Substack are not racing toward the same destination. They are building parallel futures.
Beehiiv asks: How do we help creators write better, faster, and more profitably? Substack asks: How do we help ideas travel further, in more formats, to more people?
If your work begins with words, Beehiiv’s AI feels like a well-lit writing room stocked with power tools. If your work ends in sound and motion, Substack’s AI feels like a broadcast studio that never sleeps.
In the end, the smartest creators don’t ask which AI is “better.” They ask which one makes their voice louder without making it less human.
The Enduring Resilience of Email: From Internet Pioneer to AI-Powered Essential
In a digital world addicted to speed—endless scrolls, disappearing stories, and algorithmic feeds that forget us as quickly as they find us—one technology has quietly refused to die: email.
Born before websites, before browsers, before the modern internet itself, email has not merely survived decades of technological upheaval. It has outlasted them. And in 2025, far from fading into obsolescence, email is enjoying a renaissance—powered by artificial intelligence, fueled by newsletters, and anchored by something most platforms have lost: direct human connection.
With more than 4.6 billion users worldwide and roughly 376 billion emails sent every single day, email is not a relic. It is the bedrock layer of digital communication—less flashy than social media, but infinitely more durable.
If social platforms are crowded town squares, email is still your permanent address.
Email: The Internet’s First Killer App
Long before the World Wide Web stitched hyperlinks across the globe, email quietly solved a more fundamental problem: how humans talk to each other over machines.
The earliest roots trace back to 1965, when users of shared mainframe computers could leave messages for one another. But the modern concept of networked email arrived in 1971, when Ray Tomlinson, working on ARPANET, sent the first message between computers—and introduced the now-iconic “@” symbol to separate user from host.
It was a small character with massive consequences.
By 1972, email had already gained features we still rely on today: reply, forward, and message lists. This was years before web browsers existed. Tim Berners-Lee’s first browser wouldn’t appear until 1990, yet email had already become the nervous system of early digital collaboration.
Throughout the 1970s and 1980s, email spread from military and academic institutions into corporate environments via proprietary systems. By the time the public internet emerged in the 1990s, email wasn’t a novelty—it was infrastructure.
The Web Era: Hotmail and Gmail Bring Email to the Masses
The 1990s transformed email from a technical tool into a global habit.
In 1996, Hotmail launched as the first major free web-based email service, allowing users to access their inbox from any browser, anywhere. Its clever name—HoTMaiL, a nod to HTML—symbolized email’s marriage to the web. Microsoft acquired it a year later for $400 million, accelerating email’s ubiquity.
Then came Gmail.
Launched in 2004 as an invite-only beta, Gmail felt like science fiction. While competitors offered just a few megabytes of storage, Gmail handed users 1 gigabyte—roughly 500× more than the norm—along with instant search and threaded conversations. Many assumed it was an April Fool’s joke.
It wasn’t.
By 2025, Gmail boasts roughly 1.8 billion monthly users, second only to Apple Mail. More importantly, Gmail helped redefine email not as a static inbox, but as a searchable, living archive—a personal memory bank of the digital age.
Email was no longer just communication. It was identity.
Surviving the Social Media Revolution
When social media exploded in the mid-2000s, email’s obituary was written many times.
Instead, email quietly held its ground, for reasons both psychological and structural.
Unlike social feeds:
Emails arrive directly, not via algorithms
They persist, rather than disappearing in minutes
They are owned, not rented
The numbers tell the story. In 2025:
Email marketing delivers roughly $36 in ROI for every $1 spent
Social media averages closer to $2–3
Email is 40× more effective at customer acquisition than Facebook and Twitter combined
Demographics reinforce this durability. Over 90% of adults aged 25–64 use email regularly, far surpassing social media usage in older cohorts. Social platforms generate awareness. Email drives action.
Social media is a megaphone. Email is a handshake.
Email in 2025: Bigger Than Ever
Today, email sits at the center of digital life—not despite newer technologies, but because of them.
As of 2025:
4.59 billion people use email (over half the global population)
Daily email volume exceeds 375 billion messages
The average inbox receives 80+ emails per day
58% of users check email first thing in the morning
Email’s enduring power lies in its time-agnostic richness. It supports long-form thought, attachments, contracts, personal notes, marketing campaigns, and life-changing decisions—all without demanding instant response.
In a digital ecosystem obsessed with urgency, email respects asynchrony. It lets humans think.
AI: Email’s Quiet Supercharger
Artificial intelligence has not replaced email—it has unclogged it.
Where email once threatened overload, AI turns scale into leverage. What looks like a comeback is really an amplification: email doing what it always did, but faster, safer, and more intelligently.
Newsletters: Email’s Renaissance Engine
No part of email’s resurgence is more visible than the newsletter boom.
In 2025:
25% of creators report significant newsletter revenue growth
45% expect further increases
Newsletter engagement outperforms most social channels
Interactive email elements boost engagement by 30–40%
Platforms like Beehiiv and Substack have capitalized on this shift, offering creators something social media cannot: direct ownership of audience relationships.
In an era plagued by bot traffic—where early-2025 data shows up to 70% of newsletter clicks originating from automated systems—email’s emphasis on retention, trust, and repeat engagement matters more than ever.
Newsletters are not noise. They are rituals.
Why Email Keeps Winning
Email’s resilience is not accidental. It rests on three enduring truths:
Universality – Anyone can reach anyone with an email address
Ownership – No algorithm stands between sender and receiver
Adaptability – Email evolves without breaking itself
Social platforms come and go. Messaging apps fragment audiences. But email remains the common denominator—the connective tissue linking every digital identity.
Conclusion: The Quiet Backbone of the Internet
From a single test message in 1971 to an AI-augmented global network in 2025, email has proven something rare in technology: longevity without stagnation.
It survived the birth of the web. It outlasted social media hype cycles. It absorbed AI without losing its soul.
Email doesn’t shout. It endures.
And in a digital world that grows louder every year, that quiet persistence may be its greatest strength of all.
AI’s Transformative Power in Newsletters: From Today’s Tools to Tomorrow’s Intelligent Ecosystems
In the ever-evolving digital landscape of late 2025, newsletters occupy a paradoxical position: they are one of the oldest internet-native formats, yet they feel more future-proof than most modern media channels. With more than 4.6 billion email users worldwide, newsletters are not merely surviving the algorithmic chaos of social platforms—they are thriving as direct, owned, and permission-based channels in an attention economy built on rented land.
But the true inflection point isn’t email itself. It’s artificial intelligence.
AI is rapidly infusing intelligence into every layer of the newsletter lifecycle—creation, distribution, engagement, monetization, and optimization. Today, AI acts as an efficiency engine. Within three years, it will become predictive and semi-autonomous. By 2030, newsletters may evolve into living, conversational systems—less like publications and more like ongoing dialogues between creators and audiences.
Imagine newsletters that don’t just publish content, but listen. Systems that observe opens, clicks, scroll depth, dwell time, replies, and behavioral patterns—then close the loop by reshaping future editions automatically. This article explores where we are today, where we’re headed by 2028, and what a fully intelligent newsletter ecosystem might look like by the end of the decade.
AI in Newsletters Today: Efficiency at Every Layer
As of December 2025, AI is already deeply embedded across the newsletter stack. Platforms like Beehiiv and Substack are leading the charge, integrating AI tools that compress what once took hours—or entire teams—into minutes. For many creators, AI has become the silent editor, researcher, designer, and operations assistant working behind the scenes.
Smarter Content Creation
AI writing assistants now dominate the creative layer. Beehiiv’s AI suite can generate full drafts from prompts, autocomplete paragraphs, correct grammar, shift tone (casual to professional), and expand or compress content on demand. Research assistants like Perplexity accelerate sourcing and synthesis, while workflow tools such as n8n automate repurposing across platforms.
Substack, meanwhile, leans into multimedia intelligence. Its AI-driven text-to-speech converts essays into podcasts. Video transcripts are summarized in minutes, and automatic clipping pulls short-form highlights for social distribution. Image generation tools further reduce friction—custom headers, thumbnails, and illustrations can be created in styles ranging from photorealistic to graphic novel.
In effect, AI has turned newsletters into low-friction publishing engines, where the bottleneck is no longer execution, but ideas.
Personalization and Distribution Intelligence
Beyond creation, AI increasingly governs who receives what and when. Platforms like ActiveCampaign and Encharge use behavioral data to segment audiences dynamically and predict optimal send times—boosting open rates by double digits. Built-in translation tools allow creators to publish globally without multilingual teams.
Analytics have also evolved. AI now tracks not just opens and clicks, but interaction patterns—how readers move through content, where they linger, where they drop off. These insights feed recommendation loops that suggest better subject lines, tighter intros, or restructured sections.
Some AI agents already operate at the scale of entire newsletter brands, autonomously producing daily digests, SEO-driven content, and growth loops reminiscent of Morning Brew-style operations—without human marketing teams.
The Newsletter as a Conversation
AI has also begun reshaping consumption. Long newsletters can be auto-summarized. Emails can be forwarded to agents that extract action items. AI-curated newsletters like Mindstream or Visually AI personalize daily updates based on reader preferences.
The result is measurable: AI-enhanced newsletters routinely outperform traditional campaigns, with interactivity driving significantly higher engagement. Still, today’s AI remains fundamentally assistive. It helps—but it does not yet decide.
AI in Three Years: Predictive Autonomy and Hyper-Personalization (2028)
By 2028, AI will evolve from assistant to co-pilot.
The global AI marketing market is projected to exceed $100 billion, with AI systems handling the majority of routine marketing functions. Newsletter platforms will no longer just execute instructions—they will anticipate outcomes.
Closed-Loop Intelligence
At the center of this shift will be advanced feedback loops. AI will analyze granular behavioral signals—open timing, scroll velocity, hover duration, reply sentiment—and use them to continuously refine future editions.
Churn will become predictable. AI will detect disengagement before a subscriber consciously opts out and automatically adjust content, cadence, or format to re-engage them. Newsletters will feel less like broadcasts and more like adaptive experiences.
One Newsletter, Millions of Editions
Content creation will become massively personalized. Instead of a single issue sent to everyone, AI will generate micro-editions tailored to individual readers—adjusting topics, depth, tone, and recommendations based on past behavior.
Predictive analytics will surface topics before they trend. Embedded chat interfaces could allow readers to ask follow-up questions directly within emails, turning newsletters into two-way channels rather than static artifacts.
Monetization will also become intelligent. AI will dynamically place ads, upsells, or sponsorships based on reader intent—maximizing relevance while preserving trust.
Importantly, this era will elevate ethical AI. Privacy-preserving personalization, transparent data usage, and user control will become competitive advantages, not compliance afterthoughts.
AI in Five Years: Conversational Ecosystems (2030 and Beyond)
By 2030, newsletters may cease to be documents at all.
They will become agentic ecosystems—intelligent, conversational entities that adapt in real time. Instead of “reading” a newsletter, subscribers may interact with it.
The Newsletter as an AI Companion
Imagine opening an email that asks: “Which section do you want to explore deeper?” or “Want this applied to your business?”
Embedded AI agents could generate custom explanations, simulations, or personalized frameworks on demand. Content becomes modular, interactive, and responsive.
With consent, AI might incorporate signals from wearables or cross-device behavior to optimize timing, tone, and format—creating experiences tuned not just to interests, but to context.
Predictive Creation and Co-Authorship
Creation itself will become predictive. AI will forecast audience needs based on global data, generate full issues autonomously, and treat human creators as strategic directors rather than line-by-line producers.
Static articles give way to living frameworks—content readers can customize by inputting their own data, goals, or constraints.
Monetization as Infrastructure
Newsletters may evolve into hubs for education, commerce, and community. AI-orchestrated ads, agent-led purchasing, influencer avatars, and embedded learning modules could coexist seamlessly within the inbox.
Trust will be paramount. Tamper-resistant systems, transparent AI disclosures, and user-governed data rights will define the winners in an emerging Internet of Agents.
Conclusion: The Feedback Loop That Redefines Connection
AI’s trajectory in newsletters is not linear—it is exponential.
What begins as efficiency becomes prediction. What begins as personalization becomes conversation. At every stage, AI tightens the feedback loop between creator and audience, turning raw behavioral data into insight, and insight into better experiences.
This is not about replacing creators. It’s about amplifying human voice at machine scale.
As AI matures, newsletters may become the most powerful form of owned media ever created: adaptive, intelligent, and deeply personal. The inbox of the future won’t be louder—it will be curated, conversational, and quietly brilliant.
Creators who embrace this evolution early won’t just publish newsletters. They’ll build ecosystems.
