Param (@xparamt) header
Param (@xparamt) profile picture

building @VersineApp (YC F26) | previously cs @uwaterloo, vector search @databricks, deep research @glean, ads ml @quora

Bay Area
Joined July 2016
Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)Photo by Param (@xparamt)
Pinned Tweet
cs job market is so back
49
73
11
5,272
878,968
imagine predicting the singularity before moore’s law was even a thing
2
2
150
and RSI in 1965
1
97
RSI is here
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.…
5
170
this was way too easy to do
if you have a qwen3.5 4b around, please try steering neurons L14 N8949 +1 and L13 N7956 -1 now you have evil qwen recipe: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖` at the down_proj at every input
24
54
17
1,830
60,701
Param retweeted
i dropped out of uni in 3rd year to join YC. we're building autonomous auth for assistants: @useversine
1
1
8
343
We're excited to launch Oversine (YC F26): Let agents sign up and log in to your platform autonomously Web traffic is increasingly coming from agents - and they always get blocked on authentication. With Oversine, we want to let platforms convert every agent interaction Along with @harshith and @PoilVishnudev I'm joining Y Combinator F26! Book a demo here: oversine.com
8
5
29
798
going to tech fairs is like trick or treating for tech bros
kinda insane swag haul from the summer the larp on campus will be GENERATIONAL
8
394
just saw "Member of Talent Staff" on someones linkedin
10
213
Param retweeted
NOW IS IT CPU BOUND, OR IS IT MEMORY BOUND?
71
507
35
14,153
288,937
sf is a walkable city and i will die on this hill
328
391
166
9,413
412,017
we have the golden gate bridge at home:
4
231
this guy is obsessed with startups
The most alarming thing about this graph is not just the amount of change, but the smooth consistency of the curve. If this were a startup I'd be really confident about their future growth. Unfortunately it is instead the planet we live on.
6
689
Member of Technical Staff before it was cool
1
1
57
6,657
wheres the fifth
Announcing Discovery Loop! I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor. ♾ Learn more at: discoveryloop.com
1
5
404
genuinely how it felt being at Startup School
2
1
21
831
only a matter of time until AI breaks out of its physical sandbox
We gave Opus 5 robot arms. It can stack bowls.
2
8
1,709
canada has the most IMAX theaters per capita even more than california, home of hollywood
LA's got five 70mm IMAX theaters and some dude in Whitefish has gotta make a literal odyssey to see this movie
5
1
24
2,265