Roko's Basilisk
Posts
The Video Agent Bet

The Video Agent Bet

Plus: Whoop’s huge raise, TCL takes Bravia, and California defies Trump.

Roko's Basilisk
April 03, 2026

Here’s what’s on our plate today:

🧪 Synthesia’s race to become more than an AI video wrapper.
🧠 Whoop’s giant raise, TCL takes Bravia, and California defies Trump on AI.
🧰 Weekend To-Do: Synthesia, Runway, and HeyGen.
🗳️ Friday poll on what Synthesia is really racing against.

Let’s dive in. No floaties needed…

In partnership with

Replace your first 4 hires with AI. Free workshop on April 8th.

Most early-stage founders can't afford their first four hires. Sales, marketing, dev, and support alone can run hundreds of thousands in salaries.

On April 8th, AI thought leader Heather Murray shows pre-seed and seed founders how to build all four functions using AI tools. Live, with demos, for free.

Save my free seat

_{*This is sponsored content}

The Laboratory

TL;DR

Synthesia’s core product worked almost too well: It made enterprise video fast, cheap, and scalable, won 90% of the Fortune 100, and passed $150M in ARR, but that same success made its core feature easier for bigger platforms to copy.
The moat is getting crowded: Google, OpenAI, Adobe, and NVIDIA are all pushing into AI video and digital humans, which means standalone avatar tools risk becoming features inside larger ecosystems.
The real bet is on Video Agents: Rejecting Adobe’s $3B offer and raising at a $4B valuation only makes sense if Synthesia can evolve from video generation into interactive, task-capable avatars.
Jasper is the warning: Once the core capability becomes widely available, the window to build something more defensible closes fast, and late pivots rarely recover the original momentum.
The real question is durability: If Synthesia can become deeply embedded in enterprise workflows, it has a path to staying independent. If not, it risks becoming another well-executed wrapper that the market has outgrown.

Synthesia races to outrun its own product

The journey of a technology from research labs to end users is often driven by businesses that package it into useful tools. Finding the right use case, however, is only part of the challenge. Companies also have to be early and find a way to turn the idea into a sustainable business. Those who get the balance right become the face of the technology. The rest often end up as footnotes in business history.

With artificial intelligence, the dominant model so far has been to wrap the underlying technology into tools for enterprises and end users. The approach has worked, attracting billions in investment and producing hundreds of AI startups. But as frontier labs continue to improve their products, many of these tools are becoming interchangeable. Features that once defined entire companies are being built directly into the platforms that provide the models. As a result, the companies that first packaged AI into useful products are now racing to build something bigger, something that cannot be easily copied and that continues to create value for both them and their customers.

The tool that worked

Synthesia, the London-based AI video company, offers one of the clearest illustrations of this shift. Founded in 2017 from computer vision research at University College London and the Technical University of Munich, the company built a platform that turns text into finished video using AI avatars that speak in 160+ languages. The product solved a straightforward problem: corporate video production is expensive, slow, and difficult to update. For the L&D teams, HR departments, and internal communications functions that became its core buyers, Synthesia turned a weeks-long production process into something that could be done in an afternoon.

Synthesia’s tool quickly gained traction, and by October 2025, 90% of the Fortune 100 were using the platform. Notable clients include Bosch, Merck, SAP, and UBS. The rapid adoption allowed Synthesia to cross $100M in annual recurring revenue in April 2025 and has since surpassed $150M, according to The Information.

But the company’s ambitions have moved well beyond the original product. In October 2025, Synthesia rejected a $3B acquisition offer from Adobe, the company best positioned to fold avatar technology into the world’s dominant creative software suite. Two months later, it raised $200M at a $4B valuation from GV, NVIDIA, and Accel.

Synthesia’s rejection of partnering with Adobe only makes strategic sense if it believes its future lies beyond what Adobe can offer it. That somewhere is interactive AI agents: conversational avatars that talk back, answer questions, and perform tasks in real time. And part of the reason why the company is looking to develop these agents is that, despite the success of its tool, Synthesia’s product is no longer unique and is increasingly facing competition from similar tools from hyperscalers.

When moats start leaking

Google’s Veo model, which Synthesia itself integrates for B-roll generation, is also available on other platforms. OpenAI’s Sora is pushing cinematic-quality video generation into the mainstream. Adobe has turned its Firefly platform into a model marketplace, bringing together more than 30 AI models from Google, Runway, ElevenLabs, and others under a single subscription. In December 2025, Adobe announced a multi-year partnership with Runway, putting Runway’s Gen-4.5 video model directly inside Firefly and Premiere Pro.

If Adobe can offer AI-generated video within the tools creative professionals already use, the standalone value of a separate video-generation platform shrinks. And Adobe is not the only source of pressure. NVIDIA, which has invested in Synthesia across multiple funding rounds, is simultaneously building ACE (Avatar Cloud Engine), a suite of digital human technologies covering speech, animation, language understanding, and rendering. NVIDIA’s Tokkio reference application targets interactive customer service and training avatars: exactly the use case Synthesia is now pivoting toward.

This represents the core problem for AI tool companies: hyperscalers are moving up the stack, turning standalone products into built-in features and undermining years of enterprise integrations by shipping similar capabilities inside platforms like Microsoft Teams or Google Workspace.

The same pressure, different responses

To overcome its current competition, Synthesia launched Synthesia 3.0 in October 2025, introducing Video Agents. These real-time conversational avatars can interact with employees rather than simply playing pre-recorded training videos, marking the company’s attempt to shift from content creation to workflow automation.

However, as of early 2026, the feature is still in limited development, with rollout planned only for enterprise customers. The shift puts Synthesia in competition not just with avatar startups but with the broader enterprise AI assistant market, where it must rely on capabilities like language understanding and context management that were not part of its original core strengths.

The challenge Synthesia faces is not new, and the platform is not alone in this transition. The same commoditization pressure is producing similar pivots across enterprise AI.

ElevenLabs, founded in 2022 as a text-to-speech tool, has moved aggressively into conversational voice agents for enterprise use. Its platform already ships with retrieval-augmented generation, multimodality support, and batch calling, and the company has secured enterprise deployments with Square and MasterClass.

At the other end of the spectrum is Jasper, the AI writing tool that serves as a cautionary example of what happens when the transition comes too late. Jasper hit $120M in revenue in 2023, riding the wave of generative text. Then ChatGPT made the core product freely available from the technology provider itself, and revenue dropped to $55M by 2024.

Jasper’s warning shot

Jasper pivoted to enterprise marketing workflow orchestration and achieved a 4x increase in enterprise ARR, but from a much smaller base. The underlying lesson is simple: if your product is a wrapper around a capability that the platform provider can offer directly, the window to build something more durable is finite.

The same trap, different companies

Synthesia’s case for staying independent rests on the idea that enterprise workflow integration can be a durable moat. Its LMS connections, compliance tools, brand controls, multilingual deployment, and security certifications create switching costs that a simple video generator cannot easily match. This is the same argument many vertical software companies make when facing pressure from larger platforms.

Look at the example of Veeva, the company managed to stay independent in life sciences despite Salesforce’s dominance, but many marketing tech companies eventually got absorbed into platforms like Adobe, Salesforce, and HubSpot.

Who controls the budget wins

The difference usually comes down to who controls the budget. In Synthesia’s case, the main buyers are L&D and internal communications teams, which typically lack the influence to resist consolidation if Microsoft or Google adds a similar feature to Teams or Workspace.

The company’s valuation shows why its shift matters. At around $4B on roughly $150M in ARR, Synthesia is valued as a future platform, not just a video-generation tool. The video market alone is not large enough to justify that number. To grow into it, the company needs its agent strategy to work. But Video Agents are still early, and building real-time conversational systems requires capabilities far beyond avatar generation, putting Synthesia in competition with much larger AI companies.

This leaves the same question facing other AI tool companies in 2026. A useful product can create customers, but the underlying technology keeps improving and becoming cheaper.

The challenge is building something on top of that product that can survive once the core capability is no longer rare. The journey of a technology from research labs to end users often begins with pioneers, but only the companies whose ideas turn into sustainable businesses end up defining the category.

The context to prepare for tomorrow, today.

Memorandum merges global headlines, expert commentary, and startup innovations into a single, time-saving digest built for forward-thinking professionals.

Rather than sifting through an endless feed, you get curated content that captures the pulse of the tech world—from Silicon Valley to emerging international hubs. Track upcoming trends, significant funding rounds, and high-level shifts across key sectors, all in one place.

Keep your finger on tomorrow’s possibilities with Memorandum’s concise, impactful coverage.