Chapter 3 is paste copy
Chapter 3’s an “Easter Egg” #ch3, which’s chapter 3 at my hotel in Napa California
 |
|
|
|
**Cursor’s “Composer 2” got called out as “fake” (or at least misleadingly presented) mainly because of a quick drama that unfolded right after its launch around March 19-20, 2026.**
Here’s the breakdown of what happened and why people are saying it’s “fake”:
Cursor launched Composer 2 as their shiny new **proprietary frontier coding model** — hyped as something they built/trained themselves, beating or matching top models like Claude Opus 4.6 on coding benchmarks (e.g. ~61.7% on Terminal-Bench), way cheaper (~1/5–1/10 the cost), faster, better context handling, etc. They marketed it heavily as “their” model, focused on coding data, with no mention of any base model from someone else.
**Within <24 hours**, developers spotted the real model ID leaking through the API: something like **kimi-k2p5-rl-0317-s515-fast** (or very similar variants).
→ That’s clearly a **fine-tuned / RL-post-trained version of Moonshot AI’s open-weight Kimi K2.5** (a strong Chinese open model).
Moonshot AI (the creators) initially had staff publicly confirm:
– Tokenizer matches exactly
– Behavior / outputs match Kimi K2.5
– Cursor didn’t follow the license’s attribution rule
**The license drama**: Kimi K2.5 has a clause that if your product makes >$20M/month revenue (or huge user numbers), you must **prominently display “Powered by Kimi K2.5″** (or similar) in the UI. Cursor was reportedly doing ~$100M–$167M+ monthly revenue at the time and showed zero attribution → alleged violation.
This blew up fast on X/Twitter/Reddit/HN:
– Accusations of “fake proprietary model”
– “Wrapper + little RL and pretending it’s theirs”
– Hype for $29–50B valuation raise looked deceptive
– Elon Musk and others piled on with quick jabs like “Yeah it’s just Kimi 2.5”
**What happened next (resolution)**:
Moonshot AI’s official account stepped in and basically said:
“Congrats Cursor! Kimi K2.5 provides the foundation. We’re proud. Cursor accesses it via Fireworks AI as part of an authorized commercial partnership. All good on our end.”
Cursor (via team members) also clarified they follow the license through their inference partner (Fireworks), and the post-training/RL was real (they claim ~3/4 of the compute was their own).
So the license-violation fire died down quickly once Moonshot gave the green light / partnership confirmation. But the damage to perception was done — a lot of people still call it “fake” or “dishonest marketing” because:
– Cursor never disclosed the base model upfront
– Launched with heavy “our frontier model” framing
– Got caught, only explained after community reverse-engineered it
**Bottom line today (March 21, 2026)**:
It’s **not fake in the sense of non-functional** — Composer 2 is apparently a legitimately strong coding agent (many users say it performs great).
But it **is** “fake” in the marketing/PR sense: presented as more original/proprietary than it actually was, until forced to admit the Kimi base via a leak.
Classic 2026 AI startup move — stand on the shoulders of open Chinese frontier models, add targeted RL + product integration, then hype it as “ours”. Works great until someone leaks the model string. 😅
On 09-09-39, “What They Will NEVER Teach You at Stanford Business School” debuts at 300 w 44th St at New York Fashion Week’s front row
http://www.youtube.com/watch?v=QXIaNZi3mHQ
http://www.youtube.com/watch?v=QXIaNZi3mHQ
What A Super Model Can Teach a Harvard MBA About Credit www.slideshare.net/larrychiang/what-a-super-model-can-teach-a-harvard-mba-about-credit
American Express hosts me mentoring you about FICO scores at New York Fashion Week
t.co/inxTmZAj
My video boils down 20,000 hours and moves you to the right on the entrepreneur bell curve
http://www.youtube.com/watch?v=eudADPfTWiE
Steve Jobs Texted me on 650-283-8008 in the same way that Mr Jobs called Bill Hewlett https://x.com/superSaiyanSkai/status/1941392367304761636/video/1
Larry Chiang
Fund of Founders
Founding Stanford EIR
@duck9 alum, Deeply Understood Capital Credit Chinese Knowledge 9
Solo Founder Uber API
650-566-9600 Office
650-566-9696 Direct
Cell: 415-720-8500
650-283-8008 (cell)
Editor of the widely syndicated “What They Don’t Teach at School”
whattheydontteachyouatstanfordbusinessschool.com/blog
whattheydontteachyouatstanfordbusinessschool.com/blog
CNN Video Channel: ireport.cnn.com/people/larrychiang
Read my last 10 X posts at www.X.com/LarryChiang
Author of #WTDTYASBS a NY Times Bestseller released 09-09-09 at #NYFW on a runway under the tents
whattheydontteachyouatstanfordbusinessschool.com/blog/?s=Ny+times+bestseller
www.fastcompany.com/embed/c0d4562ea2049
52 Cards. Two Jokers. What They DO Teach You at Stanford Engineering
http://www.youtube.com/watch?v=vDBY0GkI3-g
Emergency swings and cutting deals as an 9 year old
http://www.youtube.com/watch?v=OFGY7v9C4G0
Hunter Pence shared thoughts before winning WORLD SERIES’ Game #7
http://www.youtube.com/watch?v=usu0luYy9pw
{ 0 comments }
Duck9 is a credit score prep program that is like a Kaplan or Princeton Review test preparation service. We don't teach beating the SAT, but we do get you to a higher credit FICO score using secret methods that have gotten us on TV, Congress and newspaper articles. Say hi or check out some of our free resources before you pay for a thing. You can also text the CEO:
