Cross-references
Source: Google Doc · Master index for all Sutro Group meetings and logistics. See also: Meetings overview · Detailed notes · Challenge #1
Sutro Group
[General]
[Bookmarks]
[Meetings]
[Meeting #1, 19 Jan 26]
[Meeting #2, 26 Jan 26]
[Meeting #3, 02 Feb 26]
[Meeting #4, 09 Feb 26]
[Meeting #5, 16 Feb 26]
[Meeting #6, 23 Feb 26]
[Meeting #7, 02 Mar 26]
[Homework for next time]
[Meeting #8, 09 Mar 26]
[Homework for next time]
Sutro Group
[General]
[Bookmarks]
[Meetings]
[Meeting #1, 19 Jan 26]
[Meeting #2, 26 Jan 26]
[Meeting #3, 02 Feb 26]
[Meeting #4, 09 Feb 26]
[Meeting #5, 16 Feb 26]
[Meeting #6, 23 Feb 26]
[Meeting #7, 02 Mar 26]
[Homework for next time]
[Meeting #8, 09 Mar 26]
[Homework for next time]
Sutro Group
[General]
[Bookmarks]
[Meetings]
[Meeting #1, 19 Jan 26]
[Meeting #2, 26 Jan 26]
[Meeting #3, 02 Feb 26]
[Meeting #4, 09 Feb 26]
[Meeting #5, 16 Feb 26]
[Meeting #6, 23 Feb 26]
[Meeting #7, 02 Mar 26]
[Homework for next time]
[Meeting #8, 09 Mar 26]
[Homework for next Monday]
Sutro Group
[General]
[Bookmarks]
[Meetings]
[Meeting #1, 19 Jan 26]
[Meeting #2, 26 Jan 26]
[Meeting #3, 02 Feb 26]
[Meeting #4, 09 Feb 26]
[Meeting #5, 16 Feb 26]
[Meeting #6, 23 Feb 26]
[Meeting #7, 02 Mar 26]
[Homework for next time]
[Meeting #8, 09 Mar 26]
[Homework for next Monday]
Sutro Group
Mission: huge amount of energy waste due to legacy learning algorithms. Fix this.
[General]
[Bookmarks]
[Meetings]
[Meeting #1, 19 Jan 26]
[Meeting #2, 26 Jan 26]
[Meeting #3, 02 Feb 26]
[Meeting #4, 09 Feb 26]
[Meeting #5, 16 Feb 26]
[Meeting #6, 23 Feb 26]
[Meeting #7, 02 Mar 26]
[Homework for next time]
[Meeting #8, 09 Mar 26]
[Homework for next Monday]
Sutro Group
Mission: huge amount of energy waste due to legacy learning algorithms. Fix this.
[General]
[Bookmarks]
[Meetings]
[Meeting #1, 19 Jan 26]
[Meeting #2, 26 Jan 26]
[Meeting #3, 02 Feb 26]
[Meeting #4, 09 Feb 26]
[Meeting #5, 16 Feb 26]
[Meeting #6, 23 Feb 26]
[Meeting #7, 02 Mar 26]
[Homework for next time]
[Meeting #8, 09 Mar 26]
[Homework for next Monday]
[embedded image]¶
General¶
Discussion: [https://t.me/sutro_group]
Meetings: every Monday at 18:00
Locations: South Park Commons (380 Brannan Street)
Bookmarks¶
AI search -- [How to find things in Sutro group]
Meetings¶
Meeting #1, 19 Jan 26¶
Meeting in SPC main floor
Notes: [sutro meeting #1 19jan26 (energy-efficient training)]
Meeting #2, 26 Jan 26¶
Meeting in Accel board room
Hinton's Forward-Forward algorithm
notes - [Hinton's Forward-Forward Paper Discussion]
Jamie Simon implementation results [doc]
Meeting #3, 02 Feb 26¶
Meeting in SPC
Joules-measuring colab [notebook]
Barak showed his Modal workflow, Yaroslav showed his Colab workflow
Meeting #4, 09 Feb 26¶
Meeting in Palmer Square
Notes: [From_Beauty_to_Joules.pdf]
Meeting #5, 16 Feb 26¶
Notes: [Intelligence_Per_Joule.pdf]
Agenda:
Take Karpathy names: [https://github.com/karpathy/makemore/blob/master/names.txt]
the model takes random 1000 names for training from Karpathy [makemore/names.txt], and then take andom 1000 names and predicts last 3 characters for each of these names
-
Create baseline to obtain baseline accuracy and number of total operations
-
Improve baseline to reduce the number of total operations without reducing accuracy
-
Share tips/lessons
[embedded image]
Meeting #6, 23 Feb 26¶
Germaine presentation:[germaine.mp4]
Emmett's results: [doc]
Overall Meeting notes: [sutro meeting #6 meeting notes]
Meeting #7, 02 Mar 26¶
Talking about sparse parity task ([Yaroslav Sutro technical sprint #1 02mar26])
Meeting [notes]
Homework for next time¶
[sutro group homework: Drosophilla of learning]
Meeting #8, 09 Mar 26
Meeting #8, 09 Mar 26¶
Notes: [Sutro meeting #8 notes]
-
Yaroslav talked about his knowledge [sprint], and showed some [slides]
-
Yad [demo'ed] his agentic harness, encouraged people to contribute on [github]
- harness gave 1000x [https://cybertronai.github.io/SutroYaro/research/survey/] faster solution, Yaroslav [verified] correctness, [visualized] top algorithm
-
Michael showed his Claude [approach] which seemed to prefer methods from the 90s
-
Germain demos his supervisor/researcher harness, solutions seemed to prefer methods from 2010s
-
Uliana gave suggestions on the temperature to use in Germain's experiments
Homework for next Monday¶
-
Get agents to improve [sutro group challenge #1: sparse parity] (using ARD as the proxy for energy)
-
Present your results/your process/present your learnings (example, Yad's youtube [video])