ojas (@bionic_BUG) header
ojas (@bionic_BUG) profile picture

i crave knowledge and cats and borgirs || nerd, gamer, jack of not all trades || designing systems || rl envs @ @mercor

India
Joined November 2017
Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)Photo by ojas (@bionic_BUG)
future is oss 💪
Wow 🤯 I left Opus 5.5 on overnight to decompile the Linux kernel and came back to the full source code available In the future all software will be open source 🚀
1
93
ojas retweeted
breaking: guy who tries to sound profound all the time actually comes across as retarded
81
58
38
1,391
30,454
3d printing is fun
6
85
holy llm psychosis
One day everyone will understand why I was talking to AI like this when they were just kids. Btw, my advise to everyone: treat every kid nicely with love and respect. You never know.
1
159
investing in your workspace is never a waste
2
1
18
611
Optimising the vertical space
2
2
1
28
1,057
openai just released astra, now my tl is gonna be filled with ai larpers marking this day as end of everything known to mankind
2
142
Career update - Joining @mercor as Member of Technical Staff. I'll be working on RL environments, evaluations, and the infrastructure that helps train and assess frontier Al models.
9
28
2,531
ojas retweeted
the amount of compute left on the table by inefficient programming is shocking
87
51
22
1,328
74,345
ojas retweeted
Stablecoins solved payment speed. Nobody solved payment safety. Same irreversibility problem cash had, except cash never moved $100k in one wrong keystroke.
1
1
4
511
why is spending unnecessary amounts of token a flex these days
yk youre down bad when you have to use gemini 😭 @claudeai i miss u ml
1
288
If only Indian government was as obsessed with AI as it is with Ethanol blending.
150
580
38
4,090
71,168
408
3,149
251
56,746
1,602,832
ojas retweeted
we are making Flipper One (yes!!) and we'll make it your favourite little computer but we need your help
48
47
6
980
37,090
Another bro,
🚨 Socket detected malicious activity in newly published versions of node-ipc, an npm package with 822K weekly downloads. Affected versions: [email protected] [email protected] [email protected] Socket’s AI scanner flagged the malware within ~3 minutes of publication. Early analysis shows obfuscated stealer/backdoor behavior, including host fingerprinting, local file enumeration, payload wrapping, and attempted exfiltration.
3
20
2,102
Update: Socket has found 121 more compromised npm package artifacts across 84 package names, including 64 UiPath artifacts. Combined w/ TanStack, the current known total is 205 affected npm package artifacts across enterprise automation, AI/MCP, auth, workflow, and dev tooling.
🚨 BREAKING: 84 TanStack npm packages were compromised in an ongoing Mini Shai-Hulud supply chain attack, adding suspected CI credential-stealing malware. Socket flagged every malicious version within six minutes of publication. This is a developing story.
50
248
159
1,249
1,104,482
Introducing SubQ - a major breakthrough in LLM intelligence. It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA), And the first frontier model with a 12 million token context window which is: - 52x faster than FlashAttention at 1MM tokens - Less than 5% the cost of Opus Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention). Only a small fraction actually matter. @subquadratic finds and focuses only on the ones that do. That's nearly 1,000x less compute and a new way for LLMs to scale.
1,480
2,717
1,972
22,369
13,499,554
we see shit like this, and then there are non technical people pushing to production, what can go wrong?
You have to review all LLM code! Codex 5.5 tried to push this awful hack to our Metal backend when it was coding font rendering. It decided to implement hacky "robust buffer access" style OOM check inside the shader and hacked our whole Metal binding architecture to add a special bind group slot 30 (hardcoded) to deliver sizes of all buffer bindings. This of course made the binding model super slow and required extra data for each buffer.
3
150
We implemented @karpathy 's MicroGPT fully on FPGA fabric. No GPU. No PyTorch. No CPU inference loop. Just a transformer burned into hardware, generating 50,000+ tokens/sec. The model is small, but the idea is not: inference does not have to live only in software 👇
260
681
121
7,383
857,549
friendship ended with vscode, zed is my best friend now
7
107