Monit Sharma

I am starting a challenge to solve the LeetGPU problem set: the specialized "LeetCode" for CUDA and high-performance computing. I will be documenting every kernel optimization and technical breakdown in this thread. github.com/MonitSharma/pa… #GPU #CUDA #HPC #LeetGPU

Hey @IBMResearch , I have used 20 mins of QPU time (10 mins last month + 10 mins this month). What is the process of availing the 180mins offer? [ibm.com/quantum/blog/o…]

@AliGrids With extended thinking

just in case… obviously

@Kaju_Nut Happy to see my Master’s thesis advisor here.

Post-conference dinner after an excellent conference at IISER Mohali.

Problem 4: Color Inversion, implemented a 1D CUDA kernel, where each thread processes one pixel of the RGBA image. By computing the global pixel index, the kernel locates the pixel’s 4-byte group in memory, inverts the R, G, and B components. github.com/MonitSharma/pa…

Problem 3: Matrix Transpose. I implemented a 2D kernel using a 16x16 thread block strategy. By calculating global row and column indices, the kernel maps input(r, c) to output(c, r). github.com/MonitSharma/pa…

I am starting a challenge to solve the LeetGPU problem set: the specialized "LeetCode" for CUDA and high-performance computing. I will be documenting every kernel optimization and technical breakdown in this thread. github.com/MonitSharma/pa… #GPU #CUDA #HPC #LeetGPU

@Sauers_ @AnthropicAI tbh every AI company should do this, not just for Tao, but other top researchers as well.

Terence Tao "ran out of tokens" on Claude Code, causing the formalization task he was working on to fail. @AnthropicAI should probably give him free Pro lol

Shor's algorithm changed the future in 1994. In 2026, the future is still arriving one noisy gate at a time.

13/ Full writeup with code, raw hardware data, all histograms, and the two-layer analysis framework: @_monitsharma/sometimes-its-too-slow-for-shor-c01677032324" target="_blank" rel="nofollow noopener">medium.com/@_monitsharma/… First in a series. Next: applying these metrics to every published Shor result from Vandersypen 2001 to the recent N=221 claims.

Shor's algorithm changed the future in 1994. In 2026, the future is still arriving one noisy gate at a time.

1/ I ran Shor's algorithm on IBM's best (publicly available) quantum hardware and measured what actually came out. The biggest number it can genuinely factor? 15. Papers claim 91, 221. Here's why those claims don't survive honest metrics. 🧵