Love exploring contextual Technologies and music, passionate about how Javascript has gone everywhere and opened infinite opportunities for universal apps

New Delhi
Joined June 2009
Include

Only show posts containing:

Exclude

Hide posts containing:

Time range
-
Minimum likes
Vikash Bhardwaj retweeted
Anthropic has gone from ~2.5k code changes per day to 30K per day in 4 months Great piece on their building of an internal version control system called Scion anthropic.com/engineering/re…
4
4
39
4,706
Replying to @ravipal1214
Are there any details of guardrails provided which were given to Agents or these were harnessed without any rules?
1
60
Vikash Bhardwaj retweeted
Running models locally with a nuanced approach - explored different types of model categories that I have tried to explain in this article.
1
2
1
1
140
Too often, AI agents pay for tools they never use — on every single call. ~40K tokens of tool schemas, even for "hi." Lazy tool loading cut one real request from 731K → 317K tokens; 38.6% less overall across 13 production prompts. Full write-up: medium.com/@vikasbhardwaj7/t…
51
Nicely said @ravipal1214,AI can do the work, but it can’t define what “good” means. If companies don’t write down their standards, machines will learn from history — including the mistakes. Own the standard, or the standard will own you.
Nobody is going to acquire your company. But the agents doing your work are named by someone else's directory, judged by someone else's checkers, governed by someone else's gates, and pointed at goals from someone else's menu. Each rental improves the quarter. None of it appears in any register of who owns what. That is a worse outcome than being bought. At least an acquisition has a price. superpal.blog/2026/08/29/the…
3
118
Vikash Bhardwaj retweeted
The check passed. Nobody signed it. A research note about code review, three product defects, a bank's hiring plans, a stalled standards process and four pricing pages. They share one thing. In every case a system approved the work, and afterwards nobody could say which person stood behind that approval. This normally gets filed as a governance problem, which is why it normally goes nowhere. Governance problems get a policy and a working group. This one belongs to whoever owns the business case, and the arithmetic is what puts it there. superpal.blog/2026/08/23/the…
2
3
12
243
Replying to @vikash21 @kaegie
I had to implement Jira integration where I had to fetch custom fields based on ID type provided by user which require lot of data transformation for Agents to work with, since then this is first rule of thumb to always minimize token usage :)
1
17
Replying to @kaegie
Sometime people learn it by experience when they use their own subscription to pay for tokens, I recall building my first Agentic coding product in 2023...
1
1
16
Vikash Bhardwaj retweeted
Improving Deep Agents with harness engineering 👀 Our coding agent went from Top 30 to Top 5 on Terminal Bench 2.0. We only changed the harness. The goal of a harness is to mold the inherently spiky intelligence of a model for tasks we care about. Harness Engineering is about systems. You’re building tooling around the model to optimize goals like task performance, token efficiency, latency, etc. Design decisions include the system prompt, tool choice, and execution flow. But how should you change the harness to improve your agent? (teaser: self-verification & tracing with LangSmith help a lot)
16
45
12
431
42,892
Vikash Bhardwaj retweeted
Gemini 3 has impressive benchmarks for building agents (Vending-Bench 2, Terminal Bench, and Sierra). So we tested its performance as a research agent using Deep Agents. We found Gemini 3 is very effective at using research tools like file manipulation, planning, and subagent delegation. All essential for building deep agents. Follow along for a step by step guide to build a research agent using Gemini 3: youtu.be/39mZvpN0k-Q
9
35
2
221
37,341
Vikash Bhardwaj retweeted
🤖 Deep Agents JS Deep Agents is now available in JS! Written on top of LangChain and LangGraph 1.0, this brings the power of agents harnesses to the JS ecosystem Comes with planning tools, subagents, and filesystem access Try it out now: npm i deepagents Repo: github.com/langchain-ai/deep… Docs: docs.langchain.com/oss/javas…
7
43
6
289
30,507
Vikash Bhardwaj retweeted
Microsoft did it again! Building with AI agents almost never works on the first try. You spend days tweaking prompts, adding examples, hoping it gets better. Nothing systematic, just guesswork. This is exactly what Microsoft's Agent Lightning solves. It's an open-source framework that trains ANY AI agent with reinforcement learning. Works with LangChain, AutoGen, CrewAI, OpenAI SDK, or plain Python. Here's how it works: > Your agent runs normally with whatever framework you're using. Just add a lightweight agl.emit() helper or let the tracer auto-collect everything. > Agent Lightning captures every prompt, tool call, and reward. Stores them as structured events. > You pick an algorithm (RL, prompt optimization, fine-tuning). It reads the events, learns patterns, and generates improved prompts or policy weights. > The Trainer pushes updates back to your agent. Your agent gets better without you rewriting anything. The best part: you can also optimize individual agents in a multi-agent system. I have shared the link to the GitHub repo in the replies! Let me know if I should cover this in a video demo!
57
371
17
2,468
201,566
Replying to @ravipal1214
Even though Google have platforms to take benefit of this race but I feel there is going to be tough race here because others are already have upper hand at least in LLMs for speed n accuracy, will explore more on enterprise side in coming days
1
54
Vikash Bhardwaj retweeted
Subject: Complaint Regarding Encroachment and Un-authorised Use of Public Spaces in Market Corridors of Main Market, Sector 21 C, Faridabad, Haryana To The Municipal Corporation of Faridabad, Haryana, India Twitter Handle: @MCF_Faridabad Respected Sir/Madam, I am writing to bring to your urgent attention the issue of encroachment of public spaces in market corridors by shop owners, particularly restaurants, in Main Huda Market, Sector 21C Market of Faridabad, Haryana. This encroachment not only obstructs the free movement of customers and market goers but also poses significant risks to public safety. Details of the Complaint 1. Encroachment of Public Space: Most shop owners in the market have illegally occupied public corridors meant for walking by blocking the right of passage. 2. Illegally Barricading Corridors: Shopkeepers have placed permanent grills and barricades, making it impossible for customers to walk safely in designated spaces. 3. Blocking Parking Areas: Unauthorized barricades using personal grills are blocking parking spaces, forcing customers to park at a distance and causing congestion in the area. 4. Safety Hazards: Some restaurants have placed gas cylinders and heavy electrical equipment in the walking corridors, posing significant safety risks for market visitors. 5. Forcing Longer Routes: Due to the illegal barricades and grills, market visitors are forced to take inconvenient and unsafe longer routes to reach their destination shops. 6. Unauthorized Use of Public Space for Commercial Gains: Many shopkeepers are misusing public space for personal and commercial activities without permission. Expectations from MCF I request the Municipal Corporation of Faridabad to take the following immediate actions to address this issue: 1. Remove Illegal Barricading: Dismantle all unauthorized barricades and grills set up by restaurants and shopkeepers in public areas. 2. Clear Parking Areas: Remove any encroachments or unauthorized structures blocking parking spaces meant for public use. 3. Enforce Public Safety Regulations: Ensure that hazardous items such as gas cylinders and heavy electrical equipment are removed from walking corridors. 4. Enable Free Movement: Restore free and safe passage for customers and market goers in all designated areas. 5. Policy Implementation: Formulate and enforce clear policies regarding the use of rental and public spaces to prevent future encroachments. 6. Dismantle Illegal Shops: Take strict action against any unauthorized shops operating in public spaces. Supporting Evidence We urge your team to visit the market and inspect the situation. The illegal encroachments and safety hazards are visible daily and can be documented for further action. Please find attached pictures of one such restaurant taken on 29th November 2024 We believe the swift action of the Municipal Corporation will ensure a safer and more organized market environment for everyone. Thank you for your attention to this matter. We look forward to your prompt response and necessary action. Sincerely,  Citizen, Sector 21C, Faridabad #SwachhtaHiSeva #swachhfaridabad #mcfaridabad
3
11
2
24
11,525
Thank you everyone who joined us today in-person and remotely You made #ReactIndia2024 Workshop Day even more amazing 🙌 The fun isn't over yet. Join us again tomorrow for the REAL React India 2024 ⚛️🇮🇳🙏
2
5
23
4,417
Excited to share, me and @sajanjohn are speaking at @react_india to share our experiences in building AI based solutions with Edge AI and #AIagents , cant't wait to see you!! #ReactIndia2024 #publicissapient linkedin.com/posts/vikashbha…
3
5
538
Join us for a hands-on workshop where you'll learn to build and automate cutting-edge edge AI systems from scratch!
2
11
663
AI Agents are so powerful that it simplify your AI apps workflows by reducing dependency on predefined workflows and in my recent experience for Building an Agent for Jira workflows clearly see how it going to be part of every solution
1
10
Interesting article discussing the challenges that organizations face in adapting to and effectively integrating artificial intelligence technologies, specifically those related to generative AI (GenAI). #AI #genAI #agile bit.ly/3VADd1i
1
1
121