Siddhant Khare
@Siddhant_K_code
Followers
5K
Following
21K
Media
1K
Statuses
9K
23 | Devtools ❤️ | Building @ona_hq (formerly Gitpod) | Maintainer @OpenFGA, github1s 💚 | Building agent infra (TokenVM, LLMTraceFX) | views mine & working 996
Remote | India
Joined July 2018
Been tinkering with LLMs since late 2022, built some fun/random/interesting AI projects & MCP servers along the way ⚡️ Full list: https://t.co/g3X637NrPX
0
1
21
Seems like Accenture got some crazy money, last year they shake hands with Microsoft for GitHub copilot and now Anthropic for Claude Code!!!
0
0
4
The hardest problem in agentic AI isn't making agents smarter, it's making them secure at scale. We partnered with @awscloud to show how to run professional engineering agents effectively and securely inside your VPC using Amazon Bedrock. https://t.co/8HQx0c0cCu
aws.amazon.com
Discover how AWS and Ona enable enterprises to build and scale GenAI development environments with complete security and compliance. This blog explores the powerful integration of Amazon Bedrock and...
1
1
4
Crazy GPU burns today - used some RTX Pro 6000 Blackwell thanks to @CloudRiftAI for GPU credits
1
0
7
The plan was clean. The escape? Not so much. Midway City doesn’t play fair. Wishlist this 4-player coop heist FPS on Steam today.
3
9
92
Skills just landed in Codex (in experimental mode), and the significance is bigger than the diff. Skills aren’t a Codex-only invention, they’re ported from Claude Code: structured bundles of instructions (optionally scripts/resources) that the agent can discover and load only
github.com
This change prototypes support for Skills with the CLI. This is an experimental feature for internal testing.
0
0
7
Thanks to @Siddhant_K_code for the idea! https://t.co/egycGJ0TQV
I don't have enough time to spend on this rn, but you can really create your 2025 Wrapped based on your Apple Photos using my MCP server :D Please drop a comment or write me if someone does that pls :)
0
1
1
just created an MCP server to get some silly stats from your Apple Photos library about last year ✨ link in next tweet, would love to add more features, open to PRs and suggestions!
1
4
9
Took Radiology Copilot out of the lab and into a neurology room. Demoed to 7 neurologists, got thoughtful critique + praise. 4 agreed it could ease day-to-day reading and support pre-final verification. Real-world feedback hits different. Lots of great suggestions, time to
2
3
17
Took Radiology Copilot out of the lab and into a neurology room. Demoed to 7 neurologists, got thoughtful critique + praise. 4 agreed it could ease day-to-day reading and support pre-final verification. Real-world feedback hits different. Lots of great suggestions, time to
2
3
17
There’s a point in life where you stop getting bothered by things like a crucial flight getting canceled, your work getting blocked from release, or your research not getting completed. I’m glad my mentors taught me not to get too attached to outcomes, and to look for the next
1
0
10
I don't have enough time to spend on this rn, but you can really create your 2025 Wrapped based on your Apple Photos using my MCP server :D Please drop a comment or write me if someone does that pls :)
I've created an MCP server to search, organize, & analyze your photo library (like a Photo Journal) with powerful tools like Claude's brain, your iPhotos (indexes at ⚡️ speed), fuzzy search, location-based queries, and more! 🔥Examples/results below 👇🏼
1
0
5
Been playing with Claude Code subagents and there’s a subtle trap worth knowing: Subagents are separate context bubbles. They can explore, run tools, edit files, but what shows up in their tool logs (those green-dot traces you see) is mostly for you, not for the main agent. If a
0
0
4
Over the last two weeks, after a lot of retros, I’ve finally figured out what I don’t want in my life, both personally and professionally. Now I’m going to use this December hibernation time to figure out what I actually want, in terms of my role evolution and my personal life.
1
0
11
AI agents can now chain 12 (even more) API calls autonomously. Most have zero authorization checks beyond initial OAuth tokens. Wrote about why RBAC fails for agentic AI and how to build task-scoped delegation with OpenFGA 👇 https://t.co/OsljBWpZkk
dev.to
Why traditional access control fails for AI agents, and how relationship-based authorization provides...
0
3
5
AI agents can now chain 12 (even more) API calls autonomously. Most have zero authorization checks beyond initial OAuth tokens. Wrote about why RBAC fails for agentic AI and how to build task-scoped delegation with OpenFGA 👇 https://t.co/OsljBWpZkk
dev.to
Why traditional access control fails for AI agents, and how relationship-based authorization provides...
0
3
5
I’ve recorded a few videos for internal demos and How-TOs, and I’ve realized that on some days, I really enjoy making videos :D
2
0
3