“AI agents are becoming less valuable,” Lanre Basamta (CEO, Optimus AI Labs) mentioned while catching up with him during the first day of resumption.
If you are in the AI space, this statement would be tough pill to swallow, because you would ask, “Aren’t AI agents still popular in tech right now? Isn’t every startup building one? He smiled, “That’s part of the challenge, When everyone has access to the same tool, it becomes harder to stand out with it.”
What followed was a fascinating conversation that revealed his 5 AI predictions for 2026.
Prediction 1: World Models Will Emerge
“The foundation of AI is data,” he began. “But here’s something interesting: language is just one way we communicate.”
He recalled an important lesson from his diploma class back in 1999. His lecturer taught that 77% of communication is non-verbal. When you think about it, more than three-quarters of how we understand each other has nothing to do with words.
Right now, most AI tools you use, ChatGPT, Claude, Gemini, run on Large Language Models (LLMs). These models understand text, process words, and turn language into responses. But they miss quite a bit of everything else.
“Someone can dislike you without saying a word,” he said. “How would an LLM pick up on that?”
This helps explain why AI sometimes hallucinates. It reads language well, but it struggles with real life context.
It doesn’t pick up on the subtle shift in your posture when you’re uncomfortable. It can’t process the way you naturally adjust your body when you sit down in a chair. These are actions, movements, and interactions that happen outside of language.
Large language models can turn text into video, but their foundation is still language, written or spoken. They start with words and work to interpret everything else from there.
World models take a different approach because they don’t just understand language. They process video, action, interaction, and context from multiple sources.
They’re multimodal in the truest sense because these are models that can understand human actions and interactions from different sources, which is why they’re called multimodal models.
Yann LeCun, the French-American computer scientist, has suggested that LLMs alone won’t help us achieve superintelligence, but world models built on multimodal approaches might.
“2026 could be when we see the real beginning of superintelligence,” Lanre shared. “This is the year world models start to emerge.”
Prediction 2: AMI Labs Will Become a Major Player
Speaking of Yann LeCun that was mentioned earlier, there’s a startup worth watching: AMI Labs.
AMI stands for Advanced Machine Intelligence. LeCun founded it after his time as chief scientist at Meta AI, and Lanre believes this company will become one of the key players in AI this year.
Lanre believes that LeCun sees something many of us are just beginning to understand. He recognizes that language models have limitations, and he’s building something that goes beyond those boundaries.
While many companies optimize their chatbots and race to launch the next version of their LLM, AMI Labs is working on the infrastructure for what comes next. They’re building the foundation for machines that can understand the world more like humans do.
“Keep an eye on AMI Labs in 2026,” he said. “They might surprise a lot of people.”
Prediction 3: Enterprise Adoption Will Accelerate
At this point things may start getting interesting for businesses, especially in Nigeria and other emerging markets.
“Right now, AI adoption is mostly at the consumer level,” Lanre explained. “Everyone has ChatGPT on their phones. We all use Netflix, Google, TikTok. These apps have AI built in, but many companies haven’t fully adopted it yet.”
That looks set to change this year because AI has matured quite a bit. Executives have taken up courses, they’ve watched the technology stabilize, seen the case studies, and moved from curiosity to readiness. Now they’re prepared to implement.
Three years ago, business leaders wondered if AI was too experimental. Now many of them wonder if they’re falling behind competitors who are already using it.
In 2026, you’ll likely see companies integrate AI into their operations at scale. Not just a chatbot on the website, but real integration into customer service, supply chain management, financial planning, HR processes, and more.
The consumer adoption phase has been strong, now the enterprise adoption phase begins.
Prediction 4: AI Agents Will Become Commoditised
Having an AI agent appears so special or when building a custom chatbot for your business seemed like a real competitive advantage. Those days are likely coming to an end.
“AI agents are becoming commoditised,” Lanre observed. “They’re becoming so common that they’re harder to differentiate.”
You can now find single AI platforms that handle multiple tasks. Education, coding, email drafting, interview prep, social media content. One app can manage it all.
Want an AI agent that writes LinkedIn posts in your specific tone? It exists, Instagram captions? TikTok scripts? Twitter threads? Red Notes? All handled by similar tools.
This is what commoditisation looks like. When everyone has access to similar capabilities, it becomes less of a differentiator and more of a baseline expectation.
“We’ve reached a saturation point with AI agents,” he said. “Building just another AI agent without unique value is becoming less attractive.”
The market has seen an influx of AI agents as a large number of them exist now. Many do similar things, which means the novelty has faded quite a bit.
What does this mean for businesses? Instead of building your own AI agent from scratch, think about how you can use existing tools to create something genuinely different. The tool itself isn’t the innovation anymore but what you do with it is.
Prediction 5: Meta AI’s Critical Year
This prediction might spark some discussion, but Lanre shared his thoughts openly.
“Meta AI hasn’t quite lived up to expectations,” he said.
When the AI race intensified, every major tech company chose a strategy. Microsoft focused on business applications, Google targeted personal productivity, OpenAI built the models themselves, even X launched Grok but Meta chose open source.
It seemed like a smart approach at the time. While competitors guarded their technology, Meta shared theirs. Many predicted this openness would give them an advantage but things haven’t quite worked out that way.
Other players have made major progress while Meta has faced some challenges keeping pace. The company that many thought would lead this space is now working to catch up.
The twist here is that Yann LeCun left Meta to build AMI Labs, but Mark Zuckerberg brought in Alexandr Wang, the founder of Scale AI, to help fill that gap.
“Something might happen this year,” Lanre reflected. “Either they release a major product, or this could be the year that determines their position for the next five to ten years.”
Right now, the leading tier includes Google, OpenAI, Grok, and Anthropic. They’re competing with each other at the forefront. Meta is working to get back into that conversation. They’re figuring out their strategy and trying to find their competitive edge.
“2026 is when Meta either closes the gap or finds themselves further behind,” he said. “They’re around the number five position right now. This is the year they either climb back toward the top four or accept a different role in this space.”
What This Means for You
These predictions aren’t just interesting observations. If you’re building in the AI space, consider looking beyond language models alone. Exploring multimodal approaches could be valuable.
If you’re running a company, this might be your year to implement AI in meaningful ways, not just experimental ones.
If you’re investing or watching the market, AMI Labs is worth attention, and it will be interesting to see what Meta does next.
The AI space is shifting beneath our feet. Language models brought us to this point, but they might not take us to the next level alone. World models could be what moves us ahead.
2026 could be the year things shift.
