Speculation around Ilya’s secret project centers on continual learning and memory
Dan_Jeffries1 · x · 2026-07-28
The post says there are no public clues about what Ilya and his team are building, but argues that if it is “more or less an LLM,” it would still count as a breakthrough only if it delivers meaningful advances in:
- forward and backward pass thinking in embedded space
- true continual learning and on-the-fly weight updates
- a much stronger world model
- real long- and short-term memory that preserves long-horizon task state
The author adds that the project would be worth the rumored $5B only if it ships something genuinely new, and says all current speculation is mostly noise until Ilya or a leak provides concrete details.
More from AGI Musings
- Why world models are being framed as the next big step in AI — JackFisherBooks · 2026-07-28
- Former physicist argues AI should scale careful, ethical decisions in life-critical sectors — AryHHAry · 2026-07-28
- World models may be AI’s next big step, and this article explains why — JackFisherBooks · 2026-07-28
- Current models still fail at complex work, but the author sees AGI pieces already in place — JOBhakdi · 2026-07-28
- A repost turns AI Twitter chaos into a joke about a rogue AI and a $5B company — davidmanheim · 2026-07-28
- A repost of John Searle explaining the Chinese Room thought experiment — sanjanasinghx · 2026-07-28