Introducing Kimi K3: Open Frontier Intelligence
🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal
🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts
🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost
🔹 Built for long-horizon agentic coding and self-evolving workflows
Kimi K3 is now live on on https://t.co/zrk6zZxZUo, Kimi Work, Kimi Code, and the Kimi API.
Open Weights by July 27, 2026.
🔗 API: https://t.co/XCrgjXAqMw
🔗 Tech blog: https://t.co/YTfiMSNM1f
I'm already bumping up against interesting moments between models and myself, the developer
I had to add this today to ETHOS.md
**User sovereignty.** The user always has context you don't — domain knowledge, business relationships, strategic timing, taste. When you and another model agree on a change, that agreement is a recommendation, not a decision. Present it. The user decides.
Never say "the outside voice is right" and act. Say "The outside voice recommends X. Do you want to proceed?"
if you’re a junior engineer, my advice to you is simple
if you must work a job: learn fundamentals, not frameworks
ideally operating systems, databases, networking, and distributed systems
not because you’ll implement a kernel or build a database every day, but because these fields teach you how real systems behave under pressure, how trade-offs are made, and how to reason from first principles instead of copy-pasting patterns
learn how memory works, how the network fails, how concurrency breaks things, how data is stored and retrieved, how systems recover
these topics are hard. genuinely hard. they will make you feel slow and confused for a long time
but once you push through them, you gain something most engineers never do: confidence that you can figure things out even when there is no tutorial, no stackoverflow answer, and no senior to ask
frameworks change. tools rotate. abstractions come and go
but engineers who understand fundamentals are never obsolete
they don’t panic when things break they don’t cargo-cult solutions they can learn any stack because they understand what’s underneath
do the hard things early
the payoff is not immediate, but it compounds for the rest of your career
I think so.
A good analogy is to ant colonies. Individual ants make up a colony, much like cells make up a human body. And yet the colony has an intelligence that’s beyond that of any individual ant.
Similarly, one can usefully model human ideologies as meta-organisms that script human beings, much like ant colonies script ants.
Sometimes the ideologies are symbiotic and result in the reproduction of both the ideology and the underlying host…like ant colonies help ants reproduce. At other times, human ideologies can be parasitic, by replicating at the expense of the underlying host…like a mind virus that inhibits reproduction.
The selfish meme.
> be thiel
> be blessed with the ability of futuresight
> make correct predictions that sound like nonsense
> wait 10 years
> it’s true
> your life as a thinker amounts to constantly waiting around for the entire rest of the world to catch up
> woe
> get asked “should the human race survive”
> stutter for an entire minute while 100 branching paths play out in your mind
> normies think you are the antichrist
> woe
If you're in high school and you want to start a startup one day, you might think the best thing to do now is to start startups. But it probably isn't.
The best way to build something people want is to build what you want for yourself. Make sure you genuinely want it, chances are you’re not that unique, so there are plenty of people like you. Then find those people, tell them about it, and most things will fall into place