Hacker News: OsamaJaber

New comment by OsamaJaber in "DSLs Enable Reliable Use of LLMs"

OsamaJaber — Wed, 15 Jul 2026 12:18:45 +0000

The gap I've hit generating GPU kernels with agents code that compiles and runs fine but is slower than the baseline. Validator says pass, result is useless. Speed targets have to be part of the check, not just correctness

The AGI Compiler "Auto"

OsamaJaber — Mon, 13 Jul 2026 18:51:22 +0000

Article URL: https://github.com/RightNow-AI/auto

Comments URL: https://news.ycombinator.com/item?id=48897070

Points: 2

# Comments: 0

Compiler for LLMs, world models, and AGI

OsamaJaber — Thu, 09 Jul 2026 00:41:17 +0000

Article URL: https://arxiv.org/abs/2607.04542

Comments URL: https://news.ycombinator.com/item?id=48839463

Points: 2

# Comments: 0

New comment by OsamaJaber in "Auto: The AGI Compiler"

OsamaJaber — Tue, 07 Jul 2026 19:26:02 +0000

thread: https://x.com/Akashi203/status/2074495867449434389

Auto: The AGI Compiler

OsamaJaber — Tue, 07 Jul 2026 19:26:02 +0000

Article URL: https://github.com/RightNow-AI/auto

Comments URL: https://news.ycombinator.com/item?id=48822488

Points: 4

# Comments: 2

Build You Own Model

OsamaJaber — Fri, 03 Jul 2026 11:31:53 +0000

Article URL: https://runinfra.ai

Comments URL: https://news.ycombinator.com/item?id=48773786

Points: 2

# Comments: 0

RunInfra: Optimize any open model down to the kernel, deploy in 5 min

OsamaJaber — Tue, 30 Jun 2026 23:48:33 +0000

Article URL: https://runinfra.ai

Comments URL: https://news.ycombinator.com/item?id=48740717

Points: 2

# Comments: 0

Compiles any HuggingFace model into a single persistent megakernel

OsamaJaber — Wed, 17 Jun 2026 22:55:25 +0000

Article URL: https://twitter.com/Akashi203/status/2067379010762338681

Comments URL: https://news.ycombinator.com/item?id=48578154

Points: 2

# Comments: 0

Mega Kernels, Written by Agents

OsamaJaber — Fri, 12 Jun 2026 17:24:11 +0000

Article URL: https://arxiv.org/abs/2606.09682

Comments URL: https://news.ycombinator.com/item?id=48506888

Points: 2

# Comments: 0

AutoMegaKernel: Compiling a LLM into a single CUDA kernel

OsamaJaber — Tue, 09 Jun 2026 20:26:21 +0000

Article URL: https://arxiv.org/abs/2606.09682

Comments URL: https://news.ycombinator.com/item?id=48467210

Points: 3

# Comments: 0

AutoMegaKernel: Compile an LLM into one provably-correct CUDA megakernel

OsamaJaber — Mon, 08 Jun 2026 16:20:11 +0000

Article URL: https://github.com/RightNow-AI/AutoMegaKernel

Comments URL: https://news.ycombinator.com/item?id=48447329

Points: 4

# Comments: 0

StreamIndex: Memory-bounded compressed sparse attention via streaming top-k

OsamaJaber — Sat, 23 May 2026 13:11:22 +0000

Article URL: https://arxiv.org/abs/2605.02568

Comments URL: https://news.ycombinator.com/item?id=48247363

Points: 4

# Comments: 0

Show HN: AutoKernel, Auto GPU Kernel Optimization

OsamaJaber — Wed, 13 May 2026 12:13:11 +0000

Article URL: https://arxiv.org/abs/2603.21331

Comments URL: https://news.ycombinator.com/item?id=48120897

Points: 2

# Comments: 0

DeepSeek V4's indexer dies at 65K. We got it to 1M on 6GB

OsamaJaber — Tue, 12 May 2026 22:12:45 +0000

Article URL: https://arxiv.org/abs/2605.02568

Comments URL: https://news.ycombinator.com/item?id=48115294

Points: 5

# Comments: 0

AutoKernel: Autonomous GPU Kernel Optimization via Iterative Agent-Driven Search

OsamaJaber — Sat, 09 May 2026 12:39:34 +0000

Article URL: https://arxiv.org/abs/2603.21331

Comments URL: https://news.ycombinator.com/item?id=48074520

Points: 4

# Comments: 0

New comment by OsamaJaber in "A recent experience with ChatGPT 5.5 Pro"

OsamaJaber — Sat, 09 May 2026 12:35:38 +0000

the bottleneck isn't generation, it's verification

DeepSeek V4's indexer OOMs at 65K context. We got it to 1M in 6G

OsamaJaber — Tue, 05 May 2026 11:13:14 +0000

Article URL: https://arxiv.org/abs/2605.02568

Comments URL: https://news.ycombinator.com/item?id=48020854

Points: 8

# Comments: 0

Ouroboros: Dynamic Weight Generation for Recursive Transformers

OsamaJaber — Sat, 25 Apr 2026 14:15:03 +0000

Article URL: https://arxiv.org/abs/2604.02051

Comments URL: https://news.ycombinator.com/item?id=47901713

Points: 2

# Comments: 0

New comment by OsamaJaber in "Kimi vendor verifier – verify accuracy of inference providers"

OsamaJaber — Mon, 20 Apr 2026 20:04:39 +0000

Good to see this exist. Inference providers quietly swap quant levels. Most users never check. A standard verifier from the model maker is the right move, would love to see other labs ship the same

New comment by OsamaJaber in "Kimi K2.6: Advancing open-source coding"

OsamaJaber — Mon, 20 Apr 2026 20:00:37 +0000

The modified MIT clause is sneakier than people think. Hit 100M users or $20M a month and you have to slap "Kimi K2.6" on your UI. That covers any consumer app worth building. Not really open, more like free until you matter. Llama pulled the same move