The performance ceiling for non-invasive Brain-Computer Interfaces (BCIs) is rising dramatically, not from better sensors, but from advanced AI. New models can extract high-fidelity signals from noisy data collected outside the skull, potentially making surgical implants like Neuralink unnecessary for sophisticated use cases.

Challenging Neuralink's implant-based BCI, Merge Labs is creating a new paradigm using molecules, proteins, and ultrasound. This less invasive approach aims for higher bandwidth by interfacing with millions of neurons, fundamentally rethinking how to connect brains to machines.

The team obsesses over perfecting the BCI cursor, treating it as the key to user agency on a computer. However, the long-term vision is to eliminate the cursor entirely by reading user intent directly. This creates a fascinating tension of building a masterwork destined for obsolescence.

Paradromics uses LLMs to decode brain signals for speech, much like how speech-to-text cleans up audio. This allows for faster, more accurate "thought-to-text" by predicting what a user intends to say, even with imperfect neural data, and correcting errors in real-time.

Neurological studies show the human brain maps a tool's tip as if it were our hand. This implies that a powerful physical intelligence should not be tied to a specific body (e.g., a humanoid) but should be a general "brain" capable of controlling any embodiment, from a bulldozer to a multi-fingered hand.

Due to latency and model uncertainty, a BCI "click" isn't a discrete event. Neuralink designed a continuous visual ramp-up (color, depth, scale) to make the action predictable. This visual feedback allows the user to subconsciously learn and co-adapt their neural inputs, improving the model's accuracy over time.

The core technology of detecting "intent" is viewed as a platform. Once the implant is in place for stroke recovery, it can be trained to detect cognitive lapses and provide real-time prompts, creating a system to assist with conditions like dementia or MCI.

To manage expectations with patients and regulators, Epia Neuro carefully frames its device as an "assisted living solution" that helps with daily tasks for life, while acknowledging that any brain retraining benefits are currently unknown and not the primary claim.

A novel training method involves adding an auxiliary task for AI models: predicting the neural activity of a human observing the same data. This "brain-augmented" learning could force the model to adopt more human-like internal representations, improving generalization and alignment beyond what simple labels can provide.