Sabitlenmiş Tweet
Ranga Sampath | Network Expert | AI Consultant
5.1K posts
Ranga Sampath | Network Expert | AI Consultant
@youplusai
Building Agentic AI tools to solve high stakes problems in Cloud Networking. The Network Ghost Agent encodes senior architect methodology into safe-exec code.
Bangalore, India Katılım Şubat 2019
961 Takip Edilen307 Takipçiler
@johncrickett Very true and the dirt just keeps on piling
English
When did "more" become the default answer in software development?
Charles Simonyi put it perfectly:
“Programming today is the opposite of diamond mining. In diamond mining you dig up a lot of dirt to find a small bit of value. With programming you start with the value, the real intention, and bury it in a bunch of dirt.”
Sound familiar? It's most legacy codebases.
Next time you reach for a new tool, ask: am I solving a problem, or am I just adding layers?
English
𝗣𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 𝗶𝘀 𝗳𝗶𝗻𝗲.
𝘍𝘪𝘯𝘦 𝘤𝘰𝘮𝘱𝘢𝘳𝘦𝘥 𝘵𝘰 𝘸𝘩𝘢𝘵?
If there is no baseline, that question has no answer — and the post-change review has nothing to anchor the number to.
---
A token bucket filter capping all TCP to 5 Mbps returns consistent measurements every iteration. No anomaly fires. The result looks clean.
**𝗧𝗵𝗲 𝗯𝗮𝘀𝗲𝗹𝗶𝗻𝗲 𝗰𝗼𝗺𝗽𝗮𝗿𝗶𝘀𝗼𝗻 𝘀𝗵𝗼𝘄𝘀 −𝟵𝟵.𝟳𝟱%.**
That is the number *"performance is fine"* cannot provide without a before measurement.
---
🔧 The 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗣𝗶𝗽𝗲 𝗠𝗲𝘁𝗲𝗿 as a standalone tool closes this gap.
Structured pipeline — validate → preflight → measure → compute → compare → report — running qperf (latency) and iperf2 (throughput) between two Azure VMs. P90 statistics. Baseline stored in Azure Blob. Deltas computed on every run. Six JSON artifacts per run. Every command through a human-in-the-loop safety gate.
---
How can the Agentic Pipe Meter be used as a standalone network performance comparison tool @ youplusai.com/agentic-pipe-m…
English
@claudeai Can Claude attend office for me ie stand in for me at work?
English
Cloud network troubleshooting has a natural stopping point: the first clean result.
NSG clean → close the ticket. That instinct kills investigations.
🔍 A clean NSG rules out the NSG. It says nothing about routing, DNS, or anything further along the path.
Each layer must be verified independently. A clean result at one layer narrows the investigation — it does not close it.
An investigation needs to clear all the layers,
— DNS resolution
— Routing / effective route table
— NSG
— Application-layer dependency (service endpoint, certificate, IAM binding)
The most expensive failures hide in the layer nobody reached after the first came back clean. Continue until the fault is found or every layer is cleared.
Which layer have you seen investigations stop at prematurely most often?
Field notes — five themes, fourteen lessons:
youplusai.com/field-notes/
English
@dharmeshba No worries. Humans will game this very 🔜
English
A friend told me his Bangalore startup has allocated Claude Code limits for every engineer. If an engineer hits the limit faster than others, he is expected to give an explanation.
We are not losing to AI. We are losing to f**king bureaucracy.
TFTC@TFTC21
Jensen Huang: "If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed. This is no different than a chip designer who says 'I'm just going to use paper and pencil. I don't think I'm going to need any CAD tools.'"
English
Loved all the ideas that Sabari is exploring. They would make a world of difference to those who use it.
Jaspreet Bindra : Homo Promptus@j_bindra
Saturday was my last day at @AshokaUniv, where I teach a course on AI for their Young India Fellowship students. This was on vibe-coding, and one student used @GoogleAIStudio to vibe code a ‘spell-checker for @WhatsApp messages’ in the very two hours that @anujmagazine and I were teaching it! That one student was 22-year-old Sabari Venkat Sreenivas (@LinkedIn profile: linkedin.com/in/sabari-venk… ) Some students leave an impression because they have an unusual combination of courage, clarity and purpose. Sabari is one of them. Sabari is 100% visually impaired. But that is NOT the story. The story is what he has done, is doing, and what he is building. He has already lived a remarkable life for someone so young: recognised by Dr APJ Abdul Kalam & PM @narendramodi . He studied Political Science, International Studies, and is now at Ashoka. His refuses to see disability as a barrier. What really stayed with me, though, was how he spoke about AI. Many of us use AI to write , summarise, make slides, clean up documents. All good. But for Sabari, AI is not merely for productivity. It is independence. A poster can be ‘seen’; a PDF ‘read’ e; currency notes identified. Objects can be described, colours recognized, people identified. Small fish for the sighted, but for Sabari, they are dignity, autonomy and confidence. And then comes the even more impressive part. Sabari does'nt just use AI to live better; he wants to use AI to build. He is building an app called RizzVision: to make online fashion accessible for the blind. Simple things: What do I wear? What matches? Does this fit? Is this shoe aligned? Worn something inside out? Stain on my shirt? Do I look presentable for an interview? Ordinary questions for most of us, for him daily barriers to independence. Sabari is building an ‘auditory mirror’ : a way to know how you look, whether your clothing is appropriate, or something you may want to fix before stepping out. The 'mirror', a mobile app, speaks it aloud AI not just powers the product, it is impossible to do without AI Building a prototype with zero funding. Vibe-coding it with Claude using multimodal features of Google Gemini , with Chatgpt as a guide. Training on 400,000 images scrapped using AI. I have written often about the sheer possibilities AI throws up, and how ‘English becomes the new Coding’. Here AI is not just helping to consume capability but create it. AI seems to be all about hype, disruption & shiny demos. Then you meet a Sabari and remember what the best use of technology really is. Not spectacle, but agency. Not noise, but possibility. We at AI&Beyond will be standing with him to help realise the possibilities If you want to help Sabari build RizzVision with time, resources, credits, and money, please do reach out to me directly on DM. (Oh, Sabari taught me that ‘rizz’ is GenZ-speak for charisma!) @PramathSinha @ArchanaMuthappa @AshishDhawanTCF
English
𝗧𝘄𝗼 𝘀𝘆𝗺𝗽𝘁𝗼𝗺𝘀. 𝗢𝗻𝗲 𝗳𝗮𝘂𝗹𝘁 𝗳𝗼𝘂𝗻𝗱. 𝗢𝗻𝗲 𝗳𝗮𝘂𝗹𝘁 𝗺𝗶𝘀𝘀𝗲𝗱. 𝗠𝘆 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁 𝘄𝗮𝘀 𝗰𝗼𝗻𝗳𝗶𝗱𝗲𝗻𝘁 𝘁𝗵𝗲 𝘄𝗵𝗼𝗹𝗲 𝘁𝗶𝗺𝗲.
The scenario 👉 degraded throughput + complete ping failures between two VMs. The agent found a tc-tbf rate limiter on the source VM. Evidence solid. Then it closed the investigation — attributing the ping failures to "packet loss from rate limiting."
🔴 𝗪𝗿𝗼𝗻𝗴. At a protocol level, a rate limiter has no protocol awareness. It degrades all traffic proportionally. Complete ICMP failures while TCP continues is not a rate limiting signature — it is a DROP rule. The iptables ICMP DROP rule on the same VM was never investigated.
✅ 𝗧𝗵𝗲 𝗳𝗶𝘅 𝘄𝗮𝘀 𝗻𝗼𝘁 𝗮 𝗯𝗲𝘁𝘁𝗲𝗿 𝗽𝗿𝗼𝗺𝗽𝘁. It was a mandatory reasoning gate: before closing, verify — *would this mechanism alone produce this specific symptom?* That single question breaks the attribution error, regardless of how confident the agent is.
Two other things broke along the way:
→ Curve fitting: the system prompt was encoding demo solutions, not methodology
→ Linguistic guardrails: useless against confident wrong answers; they only catch hedging
Full story with diagrams and the exact gate → youplusai.com/fix-system-pro…
English
@sandislonjsak 😜 feel the same here
English
I have finally discovered the limit of Claude.
It’s me. I am the limit.
English
@casesensitiveX @ganeshsonawane @frido_official Happy to know that such a product is available in Jndia
English
English
Making movement easier with Frido Patient Transfer Lift. Designed and Made in 🇮🇳
English
𝗖𝗹𝗼𝘂𝗱 𝗻𝗲𝘁𝘄𝗼𝗿𝗸 𝗱𝗲𝗴𝗿𝗮𝗱𝗲𝗱. 𝗣𝗼𝗿𝘁𝗮𝗹: 𝗴𝗿𝗲𝗲𝗻. 𝗡𝗦𝗚: 𝗰𝗹𝗲𝗮𝗻. 𝗥𝗼𝘂𝘁𝗲 𝘁𝗮𝗯𝗹𝗲: 𝗰𝗼𝗿𝗿𝗲𝗰𝘁.
🔴 The investigation stalled — because the control plane was clean the whole time.
The answer was one command away: 𝘵𝘤 𝘲𝘥𝘪𝘴𝘤 𝘴𝘩𝘰𝘸. But OS-level inspection requires someone to decide to SSH into the VM and look. That decision only happens after you've ruled out everything the portal shows you.
🔧 The Network Ghost Agent could already investigate NSGs, route tables, effective NIC routes, and escalate to packet captures. But when the control plane came back clean, the investigation had nowhere to go.
✅ The 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗣𝗶𝗽𝗲 𝗠𝗲𝘁𝗲𝗿 integration changes that.
𝗟𝗶𝘃𝗲 𝗺𝗲𝗮𝘀𝘂𝗿𝗲𝗺𝗲𝗻𝘁 𝗿𝘂𝗻𝘀 𝗮𝘀 𝗽𝗮𝗿𝘁 𝗼𝗳 𝘁𝗵𝗲 𝗶𝗻𝘃𝗲𝘀𝘁𝗶𝗴𝗮𝘁𝗶𝗼𝗻 — 𝗻𝗼𝘁 𝗮𝘀 𝗮𝗻 𝗮𝗳𝘁𝗲𝗿𝘁𝗵𝗼𝘂𝗴𝗵𝘁. If throughput is degraded and the control plane is clean, the investigation moves to the OS layer immediately.
Azure's control-plane tools tell you what was configured. They don't tell you what the OS is doing. Those are different things.
How the Network Ghost Agent uses the Agentic Pipe Meter to run 𝗹𝗶𝘃𝗲 𝗺𝗲𝗮𝘀𝘂𝗿𝗲𝗺𝗲𝗻𝘁𝘀 𝗼𝗳 𝗹𝗮𝘁𝗲𝗻𝗰𝘆 𝗮𝗻𝗱 𝘁𝗵𝗿𝗼𝘂𝗴𝗵𝗽𝘂𝘁 @ youplusai.com/pipe-meter-int…
English
@arvidkahl Experienced the same recently -- Claude have me estimates for a bunch of things in miltiple weeks.
Then all of it got done in 3 days. Even 3 days was because of the human to give time, space and effort.
English
Claude, you'll be done with implementing ALL of these improvements in like 30 minutes, including full test coverage and security analysis.
You know it.
I know it.
Which, as Claude shows pre-AI estimates, is WILD.
English
@johncrickett github.com/ranga-sampath/…
Suite of tools to troubleshoot Cloud Networks -- all built with Claude
English
Everyone talks about how good AI agents are at writing code.
But where's the actual software?
Share your best example below.
English
@johncrickett Thank you for sharing your point of view. Yes, makes a lot of sense to sharpen and deepen the basics than spend a lot of time learning a number of additional bells and whistles.
English
@youplusai Focus on transferable skills, writing good requirements and clear thinking, defining good guardrails and tests will let you leverage any agent.
English
I spent the weekend actually reading the Claude Code docs.
It's a rabbit hole.
CLAUDE.md files. MCP configs. Skills. Subagents. Hooks. Plugins. Agent Teams.
You could spend more time configuring Claude Code than building software.
All of it is productivity theatre.
The only thing that actually matters: think first, then give it focused, relevant context.
English
@giladvdn @johncrickett Thank you for sharing your perspective. I tend towards that but have FOMO that perhaps if I had learnt a few more tricks, I would've done better and faster at the thing that I am doing.
English
@youplusai @johncrickett Create - always create. Set some time aside for learning but you'll learn a lot more from pushing a real project
English
If such an FDA approved treatment pathway is made available in India, it would certainly make the lives of so many patients seeking treatment in India, a little more easier.
Vidya Gopal@i_v_i_d
Seeking a faster regulatory path for Encorafenib (Braftovi) in India. It’s standard of care for patients with the BRAF V600 in the US and EU @Pfizer_India @CDSCO_INDIA_INF, can we expedite registration and discuss interim access programs? #PatientFirst #CDSCO
English
If anyone can help for this important cause, please do.
Vidya Gopal@i_v_i_d
This is for a stage 4 colon cancer patient in India with the BRAF v600 mutation. The Breakwater trial (Folfox + Cetuximab + Encorafenib) has shown great outcomes for this mutation. Encorafenib (Braftovi) isn’t registered in India but can be imported - though quite expensive.
English
@subhash_kak Wah Kya baat Hai...Maza Aa Gaya
हिन्दी
@toddsaunders I am ready to apply for that
English
Every company will have a Chief Claude Officer that pills new employees and keeps everyone trained / up to date.
It’s the only way and it’s that important
Sam Parr@thesamparr
How is everyone getting team adoption for Claude? I spent a lot of time on Twitter, as do you. We see all this AI stuff popping up. We're on top of it, or at least sorta. I know what's going on and are testing all these fringe ideas. But how are all you people getting your team to actually use it effectively without spending all their time on Twitter and learning, which we know they won't and probably shouldn't be?
English
@dharmesh Now, shouldn't there be a word for being emotionally down because Claude is down?
How about "lau" as in I am feeling lau!
English
Current Status: Dealing with the emotional roller coaster that is the Anthropic API right now (they're having an owie).
On the one hand, I'm in a state of flow and trying to finish up some work for a demo video I want to record a screencast for.
On the other hand, it's draining to keep retrying things 5 times in order to get it to work. Just creeping along.
My best wishes to the Anthropic team. Never fun to be going through system issues -- especially on a Friday afternoon.
English