Software Engineer @DeepSourceHQ, working with Vue.js and NuxtJS | Tinker with React, Python & .NET Core in free time | Love Gaming, Reading & Community ❤
looking for a female flatmate to take up my spot in a fully set up 2bhk in cambridge layout, halasuru ツ
- move in july 1st
- safe, quiet, residential street
- help for food/cleaning sorted
- rlly sweet n cool flatmate
dm w a short intro if interested!
@BangaloreRoomi@Flashmateshq
DeepSource AI Review now scores higher on the OpenSSF CVE Benchmark. Most importantly, we're seeing 4%+ gain on the F1 score.
This means we're catching more real vulnerabilities without introducing a single false positive.
Our precision hit 100% (zero false positives across 83 patched diffs) while recall climbed to 73.17%.
Meris is officially LIVE on Android + App Store! 🚀🔥
huge shoutout to our internal testers — your feedback carried us here.
we’re finally opening the gates — try it, use it, break it… and tell us everything.
the journey starts now → https://t.co/IxwMoKyKpm
@getmeris
We did a thing!
@hackyroot and I have been building this for a few months and I am excited to share it publicly. Give it a try, tell us how we can help you.
It is still early and we would love to hear all the feedback you have 💖
https://t.co/jcVU8wHmMz
New: REST API for Agentic Code Security 🤖🔒
Bring production-grade AI code security into your app, agent, or CI in minutes.
Read the full announcement: https://t.co/PVH4m9y4mD
@stonecharioteer@Scmmishra U4Ts are my favourite tactile switches, hehe.
The benefit of doing things in a custom way, is that I have been going super light with the spring weights in my switches. It has brought 10x improvement in my day to day.
@Faraaz98@stonecharioteer@Scmmishra For local options, a little timing is everything. The F1 8X V2 was an amazing deal and ran sometime back, if TKLs are your thing.
Stacks, Neo macro, etc. bring some very solid boards in group buy format.
Mine are mostly imported though, things have become much better recently.
@Scmmishra@stonecharioteer The number of options lately has exploded. Pretty sure, with a little timing, you should be able to find a similar layout, and even a low profile one if you prefer that. That should keep the switch minimal. 👀
@stonecharioteer@Scmmishra I just prefer linears overall. I have a couple of tactile builds, but found them to be more fatiguing at the end of the day.
@stonecharioteer@Scmmishra We should XD I keep rotating my keyboards, but my current favourites are the Kohaku, Orbit and Hiro. Mostly prefer 65%s!
@Scmmishra let's build a keyboard! 👀 Lots of local options in wireless as well now.
🤖 Introducing Narada -- a hybrid secret detection system that pairs regex for rapid scanning with a fine-tuned Llama3.2-3B-Instruct model, achieving 97% precision.
Benchmarks:
- 97% precision in secret detection
- 92% reduction in false positives (vs base model)
- 96.7% F1 score (vs regex-only tools)
Developers often leak secrets like API keys into repos, and traditional scanners drown teams in false positives due to missing context. Pure heavyweight LLMs are powerful but too slow and expensive when scanning frequently at scale.
Narada fixes this with a hybrid approach: Fast regex for candidate spotting, plus a fine-tuned Llama3.2-3B-Instruct for contextual verification. Outcome: High precision at a fraction of the cost, handling thousands of files without slowing down your pipelines.
Integrated into @AutofixBot. More details in our deep dive: https://t.co/scbJbNUd5g
Narada-3.2-3B-v1 is open on Hugging Face: https://t.co/bAQU8D5pDH
Early partners, shoot us a DM to integrate! Dropping soon in TUI/MCP 🚀
🤖 Announcing Autofix Bot -- AI agent that detects and fixes security vulnerabilities in code.
Autofix Bot v1 benchmark:
- 8% more accurate than Claude Code
- 52% cheaper than Codex CLI
- 5x faster than Gemini CLI
Programming with AI introduces real challenges to security:
- LLMs often miss critical flaws when used for self-review, with recall often dropping to zero.
- AI codegen tools produce code with hidden security risks that aren't obvious at first glance.
- LLMs use patterns from old data, introducing outdated vulnerabilities.
Autofix Bot solves these by combining AI with static analysis for consistent, deterministic results: it scans pull requests for security issues and suggests in-place fixes to prevent security issues from reaching production.
Benchmarked on OWASP's rigorous dataset of 2740 labeled files, Autofix Bot delivers 88% accuracy and 94% true positives -- matching OpenAI Codex while outperforming Claude Code and Gemini CLI in cost (58 USD vs. up to 300 USD) and time (2.4 hours). For secrets detection, it achieves a 0.93 F1 score, higher than just relying on tools like Gitleaks at 0.76, leading to fewer missed credentials and minimal false positives.
In practice, fixes average under 13 seconds, and it scales to thousands of files without slowing workflows.
Early access is open for partners -- reach out to integrate the API. Coming soon for individual developers. More information at https://t.co/H9ZW8wcrGM
Hello, world! 🤖
We're excited to announce Autofix Bot today — an AI agent purpose-built for code security. On OWASP Benchmarks, Autofix Bot performs on par or better than tools like OpenAI Codex and Claude Code, and is at least 60% cheaper.
Read the full benchmark results ⬇️
just launched Vitae 🚀
it’s an MVP — a bit buggy, but live.
lets you make resumes + cover letters with AI.
would love if you try it out & give feedback ❤️
👉 https://t.co/uE4QDN0ea0