HackerRank @hackerrank

Change the world to value skills over pedigree. hackerrank.com Mountain View, CA Joined June 2012

Tweets

14K
Followers

97K
Following

3K
Likes

12K

HackerRank @hackerrank

14 hours ago

More MCP tools will just make your agent worse. When your MCP has hundreds of tools to choose from it becomes a liability. The model has to read every tool definition. Choose the right one, pass the right inputs, and not blow up the context window before the work even starts. The better pattern isn't giving the agent more choices. It's giving it a cleaner way to find the right one. Instead of giving the agent hundreds of preloaded tools, give it a smaller interface: search what exists, write the code, then execute the right call.

1 0 7 608 2

View Details

HackerRank @hackerrank

2 days ago

Behind every great team is someone who made the harder, better call. Someone who looked past the resume and hired for what people could actually do. The HackerRank Innovator Awards is the night that work gets recognized. 📍 The Ritz-Carlton, Bengaluru | June 12, 2026 Register now: hackerrank.com/events/innovat…

0 1 6 2K 2

View Details

HackerRank @hackerrank

3 days ago

@apexinsideapex Pretty cool!

0 0 1 74 0

View Details

HackerRank @hackerrank

3 days ago

The LeetCode round in an interview tested one skill: writing code. AI does that easily now. So we rebuilt the technical interview across the task, the evaluation, and the experience. Meet HackerRank 2.0.

4 11 88 5K 51

View Details

HackerRank @hackerrank

3 days ago

Old experience: clunky IDE with no AI. New experience: agentic IDE with the latest models.

0 0 0 182 0

View Details

HackerRank @hackerrank

3 days ago

Old evaluation: can this candidate write functionally correct code? New evaluation: can the candidate orchestrate AI? We measure AI fluency, not just pass/fail.

1 0 5 1K 1

View Details

HackerRank @hackerrank

3 days ago

Old interview task: solve this LeetCode problem. New interview task: solve real problems in actual code repositories.

1 0 5 2K 3

View Details

HackerRank @hackerrank

a week ago

Developer wakes up from a coma after 2 years

1 14 124 16K 35

View Details

HackerRank @hackerrank

a week ago

Vibe Coders using Claude Opus 4.7 Max to write a hello world program

0 4 52 6K 4

View Details

rvivek @rvivek

a week ago

This is a company of 2026. Adapt or be left behind.

2 2 27 2K 3

View Details

HackerRank @hackerrank

a week ago

Agents fail in production not because the model is bad, but because nobody built the harness. Most teams focus on prompts and context windows. The reliability problem is somewhere else. Here's what actually makes agents production-ready: > Instructions should be treated as routing logic, not documentation. For humans, documentation is reference. For agents, they’re routing signals: what matters, what order, what to prioritize. > AI raises the cost of ignoring discipline. Agents become confidently wrong with dirty state, flaky setup, and unclear scope. Fundamentals matter more now, not less. > Agents don’t know what "done" is. Developers should externalize it: acceptance criteria, verification gates, clean repository markers. > Good harnesses shape behavior, not just catch mistakes. Scope controls we fix reduce over-editing, and verification gates force evidence and end-to-end checks that make agents reason about outcomes, not just code. > Sessions aren’t isolated work. Agents lose intent faster than code. Without handoff artifacts and decision logs, every new session rebuilds from scratch. Building the harness is the real engineering skill now.

0 0 11 1K 12

View Details

HackerRank @hackerrank

a week ago

Campus hiring teams are spending hours manually checking whether candidates followed the rules. We built something that enforces those rules by default. Bringing together university recruiting leaders in Hyderabad for a small, hands-on session: → Live Desktop App demo → Peer discussions on integrity at scale → Try it yourself Second stop on our India roadshow. 📍 29th May, The Leela Hyderabad Register now: luma.com/4qmpg1on

1 1 25 2K 7

View Details

HackerRank @hackerrank

a week ago

Most browser agents don't fail because they can't reason. They fail because they click the wrong thing. Anthropic's browser-agent guide breaks down why. Most of the work is at the interface layer. Four fixes that matter: 1. Keep screenshots and coordinates in sync If the screenshot gets resized before the model sees it, every coordinate can shift. The model may choose the right button and still click the wrong place. 2. Use deep reasoning selectively Planning, recovery, and unexpected UI states need deeper thinking. Routine clicks don't. Save the tokens for the moments that matter. 3. Manage screenshot bloat Every browser action can add ~1K to 1.8K tokens of screenshot context. Long tasks need pruning, rolling buffers, and cache breakpoints. 4. Batch carefully Batching saves time until step one fails. Then every next action runs on the wrong screen state. Reliable browser agents aren’t built by tuning the prompt alone. They’re built by fixing the control loop around the model.

1 0 1 875 0

View Details

HackerRank @hackerrank

a week ago

We're hiring an AI-native storyteller. If you can write well, know AI inside out, and are allergic to AI slop, here's the link to apply: hackerrank.com/careers/job?id…