I founded Crowdbotics & LeadGenius. ML, social enterprise, crowd computing, polytopes, startups. Sometime Berkeley PhD + YC alum.
Berkeley, California
Joined September 2009
- Tweets2.4K
- Following527
- Followers1.5K
- Likes1.5K
anand kulkarni 🇺🇸 retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
anand kulkarni 🇺🇸 retweeted
This past year, Nabila Hassan used her earnings as a LeadGenius researcher to build a borehole well in her home village of Bomani, #Kenya that provides more than 500 families with safe drinking water. hubs.ly/H09zlGf0
anand kulkarni 🇺🇸 retweeted
visualizing individual word vectors as circular blobs (distance of each point on the perimeter proportional to the value of a dimension)
anand kulkarni 🇺🇸 retweeted
Good guide on building your own neural conversational agent.
blog.statsbot.co/chatbots-ma…
anand kulkarni 🇺🇸 retweeted
Schwarzenegger is calling for gerrymandering reform in all 50 states, funding legal challenge to WI district maps. sfchronicle.com/politics/art…
anand kulkarni 🇺🇸 retweeted
Common mistakes when newcomers use Neural Networks. Good to avoid all of these. theorangeduck.com/page/neura…
anand kulkarni 🇺🇸 retweeted
I recently excavated a forgotten computer story generator from the early 1960s—my paper on the system is out now at: researchgate.net/publication… – at Santa Cruz, CA
anand kulkarni 🇺🇸 retweeted
What a fraternity hazing death revealed about the painful search for an Asian-American identity nyti.ms/2ft9oxH
anand kulkarni 🇺🇸 retweeted
Thinking about the rewards, material & otherwise, of pre-Seed & Seed investing beepartners.vc/seed-investme… #investing #startup #venturecapital
anand kulkarni 🇺🇸 retweeted
Diamond, an open-source Python solver for certain kinds of generalized linear models multithreaded.stitchfix.com/… via @stitchfix_algo
anand kulkarni 🇺🇸 retweeted
The Surprising Things Statistics Tell Us About Fiction newyorker.com/books/page-tur…
anand kulkarni 🇺🇸 retweeted
The "Facebook AI had to be shut down" story is clickbait "journalism", devoid of substance. Pure hype & false. Author should be ashamed.
anand kulkarni 🇺🇸 retweeted
@polybot temporarily left stealth mode to share with us his new project: crowd-sourced software theaccidentalengineer.com/cr… – at California, USA
anand kulkarni 🇺🇸 retweeted
Nice coverage of Orchestra/Flash Orgs/research from @dretelny @thisisdhaas @lydiagu @polybot @msbernst @stanfordmav! qz.com/1039981/robots-are-re…
anand kulkarni 🇺🇸 retweeted
Software is still too hard to make. Imagine when whatever is the equivalent of camcorder::movies happens for making software.
anand kulkarni 🇺🇸 retweeted
One neural network that simultaneously learns image classification, captioning, translation, speech rec. & others arxiv.org/pdf/1706.05137.pdf
anand kulkarni 🇺🇸 retweeted
Wow. Nearly 50% of the $2B raised on GoFundMe in 2016-2017 was for medical expenses: bloomberg.com/news/articles/… – at Alexandria, VA
