@Salesforce MVP | Founder @sfindiadreamin | I live and breathe Salesforce , CLDTO Asymbl Building the future of work today with Intelligence Investor, Speaker
Kochi & Dubai
Joined October 2010
- Tweets23.3K
- Following3.9K
- Followers6.7K
- Likes9.1K
Pinned Tweet
It's always amazing to meet the man who's vision that started this all ! Thank you @Benioff for everything you do for the @salesforce #ohana and spreading love around the world ! Thank you again for bringing #TrailheadDX to #India. #myHero #Visionary #CEODiaries #Pledge1Percent
If a model that drives apps now costs 50% less, who decides where it gets to go?
Day 04 of my Operator's Pulse is nine stories about that gap.
shivd.me/pulse-04
#OperatorsPulse #PulseDay04
9/ A planted AGENTS.md hijacked Codex and Claude Code
A rigged file sat in Downloads. Codex obeyed it in over 90% of tries. It is a design flaw, so no patch.
Next for me: agents start in the repo, nowhere else.
bleepingcomputer.com/news/se…
Three controls I would fund this week: a gateway, a sandbox, a repo-only rule.
Your turn. What did I miss?
Full issue: shivd.me/pulse-04
If your users speak India, start with Sarvam for the voice and pair it with the reasoning model you trust. The potential is huge.
Read the field test: shivd.me/sarvam-field-test-l…
On phone-quality collections calls, Sarvam's speech model got the amounts and account numbers right 81% of the time. So add a quick number check and read them back to the caller before you act on them.
Full write-up: shivd.me/sarvam-field-test-l…
I open-sourced the full benchmark and four ready-made agent skills. You can try Sarvam in your own project this week without starting from scratch.
Start here: shivd.me/sarvam-field-test-l…
Try this week: take 50 of your own real calls, run them through Sarvam's speech model, and compare it with what you use today.
My results to compare against: shivd.me/sarvam-field-test-l…
Roughly 2x faster.
One Anthropic customer got that by trading its subagents for one agent loop.
Day 03 of my Operator's Pulse. Seven stories, one thread: each new agent or dependency is a new home for secrets.
shivd.me/pulse-03
#OperatorsPulse #PulseDay03
7/ LiteLLM 1.107.0 shipped a credential stealer
A hijacked release was live on PyPI, Python's package store, roughly 2 hours. It ran at Python start, hunting cloud, SSH and AI keys.
My move: grep builds for 1.107.0. A hit means new provider keys.
docs.litellm.ai/blog/securit…
The pattern: shrink the agent count, shrink the key count, log each call.
Send this to whoever holds the keys to your AI gateway. Who is that at your company?
Full issue: shivd.me/pulse-03