Michael D. Moffitt (@mmoffitt) header
Michael D. Moffitt (@mmoffitt) profile picture

Member of Technical Staff @ Ndea Previously: Google DeepMind

Austin, TX
Joined July 2011
Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)Photo by Michael D. Moffitt (@mmoffitt)
Michael D. Moffitt retweeted
New #kaggle competition "Santa 2026 - The Elf Calculator" is launched. Medal: True Kernel Only: False Deadline: 2027-01-28 23:59:00+00:00 kaggle.com/competitions/sant…
10
2
75
4,975
Michael D. Moffitt retweeted
Want to try weight-space programming? Today we're releasing dspy-wsp, a prototype weight-space optimizer that works like any other @DSPyOSS optimizer. Using dspy-wsp, and about ~$5-10 of Astra, we improved every decision model we tested. cmpnd.ai/blog/weight-space-p…
29
78
13
665
38,075
Michael D. Moffitt retweeted
In Nov 2016, my team and I published a paper on Neuro-Symbolic Program Synthesis. This was a time before transformers and LLMs, and the programs we generated with neural networks were tiny macros used to perform simple data edits (think autocomplete) in sheets. Fast forward 10 years: we used LLMs to generate over 6 million lines of Lean mathematical proofs to formally verify key security properties of Signal's cryptographic protocol, all the way down to the machine-code level. These advances have been transformational. The next 10 years are going to be important and I hope we all appreciate the responsibility that comes with leveraging these capabilities for the benefit of humanity. Read more: Nov 2016: arxiv.org/abs/1611.01855 Nov 2026: deepmind.google/documents/36…
23
44
6
513
19,581
Michael D. Moffitt retweeted
New open role: Research Platform Engineer. Help define how infrastructure works in the age of agentic research engineering. Build tools for humans and agents. Also: $10k referral bonus. Details: ndea.com/jobs/research-platf…
4
2
1
16
4,028
Michael D. Moffitt retweeted
This week, I had an opportunity I could not reject: Presenting my team’s 2nd place prize-winning solution to Kaggle’s NeuroGolf 2026 competition IN-PERSON at IJCAI 2026 in Bremen, Germany. I have never presented at ANY talk or conference before, let alone a well-regarded one (try NeurIPS next? 😂 ). This went far above the goal I set at the beginning of 2025. Stage fright? Not at all Need to memorize a script? Not at all I trusted that I knew the material well enough. I poured my hearts out and made a large portion (some unpopular) of the design decisions for the team during the competition. I knew the choices we made, what worked, what didn’t, from the inside out and that’s enough to be confident and speak naturally. I am prepared for any question that may come out. I was right, it turns out my historical fears of public speaking were totally inexistent. From the beginning of NeuroGolf, I badly wanted to place in the top 3 to get this opportunity. I posted on social media, reached out to the top 10 teams via email for potential team ups, only 1 agreed. Later, we received over 30 different requests to team up but we remained extremely selective so each new addition compounds our team's capabilities. I have been burned too many times by pursuing my own interests in “stealth” while going the common route, the one my family thought was best suited for me, as my primary responsibility. It just won’t work. I need to ask for what I want even if it looked unrealistic. I cannot expect anyone to read my mind. The worst I can receive is a “no”, and even that has an opportunity to turn into a “yes” someday. And I made it here! My Kaggle experience has taught me: I just don’t get much joy from simply executing someone else’s solution. I want to be good enough to create my own! I would love to thank @mmoffitt and @ChrisDeotte for the lunch! Chris, you have helped me so much in my Kaggle journey! I am so grateful! Without you generous sharing, I don't think I would have progressed as much! It's a pleasure to finally get to meet you in person! I still remember my "genesis" on Kaggle, forking your "How To Train Open Book Model - Part 1" training notebook!
2
4
16
820
Michael D. Moffitt retweeted
It can always get worse gif
13
179
4
5,835
145,539
Michael D. Moffitt retweeted
1
1
11
337
Michael D. Moffitt retweeted
I will be speaking on our 2nd place Kaggle NeuroGolf competition solution! 18 August 2026  Tuesday 16:35- 18:00   Room 4.1 A with a capacity of max 220 participants. see you there! #IJCAI2026
Just reached Bremen! Excited for the conference!
2
3
1
14
2,167
Michael D. Moffitt retweeted
First time we figured out any reasoning method with neural networks: - AI progress moves decades forward - new trillion dollar companies started popping out almost overnight - all exams and competitions got solved by AI - any notion of cyber safety gets shattered Discovering new, different, more efficient method of reasoning does not seem impossible…
39
31
2
717
37,634
Michael D. Moffitt retweeted
Two podium finishes for NVIDIA’s Kaggle Grandmasters. 🏆 🥇 1st — 2026 NeuroGolf Championship 🥈 2nd — KDD Cup 2026 Data Agents Competition In NeuroGolf, the team used manager agents to coordinate specialized workers and build a shared knowledge base with reusable tools. For KDD, their purpose-built harness narrowed the performance gap between a frontier LLM and a smaller open model.
2
22
2
172
20,166
Michael D. Moffitt retweeted
In the recently completed NeuroGolf competition, we finished 2nd / 2963 teams. Finally, after 3 years, I broke into the top 100 in Kaggle Competitions and earned my 4th gold medal! It took me 2 years after my first win in the PII competition in April 2024 to become a prize winner again! I had thoughts "was I even improving?" the answer is, just don't give up! Progress sometimes isn't a straight line. We had a diverse team, Jacek and Yuichiro are both experienced heuristic golfers while me, @chankhavu and Niu Jin were more experienced on the ML side! Each team member ran their own pipeline independently without much sharing, to improve the diversity. I think this was crucial for us avoiding local optimas!
7
3
48
4,410
Michael D. Moffitt retweeted
it is pretty funny showing language models the Jacobian disproof and blowing their minds over and over again
51
560
29
9,717
340,865
Michael D. Moffitt retweeted
I’ve finally earned the title of Kaggle Competitions Grandmaster!
10
5
142
3,797
Michael D. Moffitt retweeted
や〜っと NeuroGolf のメダルが反映されて、2枚目の銀メダル🥈を獲得しました! あとは金メダルを目指すのみです🥇 また、チームを組んでくださった akihiro さん、k256 さん、Expert 昇格おめでとうございます✨ #Kaggle
1
3
30
1,465
Michael D. Moffitt retweeted
I received a Competition Gold Medal in the The 2026 NeuroGolf Championship Competition! Woohoo! #kaggle #goldmedal - kaggle.com/certification/com…
2
4
54
2,211
Michael D. Moffitt retweeted
hello there the jacobian conjecture is false thanx to my close friend akhil for asking about it and my other close friend fable for working during the world cup final ((1+xy)^3 z + y^2 (1+xy) (4+3xy), y + 3 x (1+xy)^2 z + 3 x y^2 (4+3xy), 2 x - 3 x^2 y - x^3 z): \C^3\to \C^3, has jacobian determinant -2, and sends (0, 0, -1/4), (1, -3/2, 13/2), and (-1, 3/2, 13/2) to (-1/4, 0, 0)
1,694
5,364
3,152
43,723
40,117,403
2nd place on Kaggle's NeuroGolf challenge 🥳 I was blessed with incredibly strong teammates! Meanwhile, International Math Olympiad (IMO) 2026 just started. Hopefully I will have something cool enough to share... 👀
2
4
31
3,961