shivangi

CTO: We lost our strongest backend engineer today. Founder: The one handling infra and outages? CTO: Yes. Founder: Did a bigger company hire him? CTO: No. Founder: Then why quit? CTO: He said he was exhausted. Founder: From the workload? CTO: Not exactly. From watching the same database bottleneck, same queue lag, same deployment mistakes come back every month. Founder: That happens in fast moving teams. CTO: He agreed. What he could not accept was that every fix was temporary because nobody wanted to slow down and clean the system properly. Founder: We had deadlines. CTO: He had standards. Founder: So he left because the work was hard? CTO: No. He left because he was not doing engineering anymore. He was just containing damage. The best engineers do not hate hard problems. They hate preventable problems that management keeps normalizing.

Seeing some of my brilliant friends fail FAANG interviews makes me believe that rigid processes and fixed rules can sometimes overlook great talent.

@Siddhant_K_code These topics are the need of the hour for enterprises adopting AI

KubeCon India 2026 Committee thinks these topics are not worthy of a stage. I'm happy to give a talk on any of these at an APAC / EU conference. If not, I will publish a YT Video or something in a blog form. IMO, it is important for enterprises. Pls correct me if I am wrong

@Siddhant_K_code Very useful and insightful, the amount of tokens and resources that gets wasted for even smallest tasks not to mention time. Persisting context becomes very crucial at enterprise level

This might be interesting for teams running agents on the same codebase every day. Most agents start every session from scratch. No memory of yesterday. No knowledge of what other agents learned. No awareness of past mistakes. A team of ten engineers running five agent sessions a day generates fifty sessions of institutional knowledge daily. And throws all of it away. Three types of memory that change this: 1. Session memory. The conversation history within a single run. Simple, but it grows with every step. By turn 20, you're sending 200K tokens per turn. The cost grows quadratically, not linearly. 2. Persistent memory. Survives across sessions. When an agent finishes, it saves what it did, what it learned, what went wrong. Next time, it loads those summaries instead of rediscovering everything. The simplest version is an AGENTS.md file. The sophisticated version uses a vector database. 3. Shared memory. One agent's knowledge available to others. The code review agent discovers a tricky initialization sequence. The code generation agent working on the same module should know about it. Without shared memory, every agent is a new hire on their first day. The most valuable form: learning from mistakes. When a human corrects agent output, that correction is signal. Store it. Retrieve it next time. An agent that repeats the same mistake twice is a tool problem. An agent that repeats it once is a memory problem. I wrote about this in Chapter 19 of the Agentic Engineering Guide

The part of being the strong one no one warns you about. When you finally admit you’re barely holding it together, people still say, “You’ve got this.” Not because you do. Because they need you to.

Thank you @ramkumarkb for sponsoring me on @github. You can join them at my sponsors profile: github.com/sponsors/Siddh…

Good stuff, nice addition to have for your ai assisted code setup

🤯 I subscribed to 30+ AI newsletters and still missed what mattered for my work. So I built a side project - skillfeed.dev - a daily brief that's different for every reader. 🚀

@AseemShrey Great Stuff 🙌

I made Claude and Codex argue until my code plan was actually good. ⚡️ Built a Claude Code skill where Codex reviews Claude's plans. They go back-and-forth until Codex approves. 🎯 3 rounds. 14 issues caught. Zero manual review. Skill & Blog : aseemshrey.in/blog/claude-co…

RT this + Signup on aigrants.in = Lifetime perks + Unlimited tokens this @OfficialINDIAai week thx to donors

Researchers Find 175,000 Publicly Exposed Ollama AI Servers Across 130 Countries thehackernews.com/2026/01/resear… via @TheHackersNews

@_nancychauhan Congratulations @_nancychauhan 💜

Keep building. Keep showing up 💪 Grateful to be in the top 3% of most active speakers in 2025 🙏 Thankful for all my amazing co-speakers, I learned so much from each of you.

Hey folks! Is anyone in Bangalore willing to foster two cats for few months ? Please let me know or DM me. 💛🐾

What's the easiest way to hack into your system? 🤔 That's the first question attackers ask. And it's exactly what security teams try to answer with attack trees. I was experimenting with threat modeling this weekend and ended up building a tool 🛠️

Distill playground is live - Paste your own chunks or use sample data - Real API with OpenAI embeddings - See clustering + deduplication in action Running on @render free tier - first load wakes the server (~30s), then ~500ms-2s incl. embedding gen. distill.siddhantkhare.com/playground

In dysfunctional families, the one who names the unhealthy patterns becomes the “difficult” one. Why? Because they disrupt long-standing generational trauma. So if you’re the blacksheep in your family, I know how much it hurts to be rejected for standing up for yourself and your values. But you’re not the problem. Hold your ground. You’re the cycle-breaker.

2025 Wrap! 🎓 Graduated from @Cornell (Master’s program) 🎙️Gave 10+ talks across conferences, meetups, and online events (London, US, virtual) on AI, Observability, Sustainability, and MCPs!