Sparsh Goyal

Sparsh GoyalMay 26, 2026

The first candidate in the panel decides everyone else's score. Nobody admits this. Everyone does it. We interviewed 6 people for a senior role last month. Same panel. Same questions. Same rubric. First candidate was a 7. Solid. Not amazing. In the debrief, someone said "she's the bar." That sentence killed the next 5 candidates. Candidate 2 was actually stronger. Specific examples, sharper tradeoffs. Got an 8 in my notes. In the room? "She's good but not that different from the first one. Pass." Candidate 4 was a clear hire. The kind of person who'd outperform within a quarter. In the room? "I liked her, but I keep comparing her to candidate 1 and I'm not sure she's that much better." We hired candidate 1. Not because she was the best. Because she went first. Every candidate after her wasn't being measured against the rubric anymore. They were being measured against a memory of a person who happened to walk in the room before them. We didn't run 6 interviews. We ran 1 interview and 5 comparisons. And we called it a process.

Sparsh GoyalJun 1, 2026

The "I already read your resume" lie. My panel lead said it last week. Three times. To three different candidates. He had not read the resume. I watched him glance at it for 4 minutes before each call. Saw the last company. Saw the title. Closed the tab. Opened Zoom. Then he'd say: "I've read your resume, so let's skip the basics." What he meant: I have no idea what's on this resume. His first question: "Tell me about your last role." She had two roles in the last 2 years. Both on page 1. The relevant one was the second. He didn't know that. She told him about the first one for 12 minutes. His second question: "What's the most complex system you've designed?" It was the lead project on page 1. Three bullets described it. She walked through it again, in the same order. 9 more minutes gone. Final question: "Why are you looking to move?" A 2-sentence summary at the bottom of the cover note answered this. He hadn't opened the cover note. Out of 45 minutes, he had 11 left for real questions. He asked one technical follow-up and ended with "this was great, we'll be in touch." Scorecard the next morning: "Technical depth: hard to assess in the time available. Recommendation: another round." We added another round. To her calendar. Because he hadn't done the 4 minutes of prep that would have made the first one useful. Here is what nobody on my team says out loud. It isn't just him. Five panelists. None read the resume. All open with "so I've reviewed your background." All burn the first 15 minutes asking her to summarize the PDF on their screen. She explains it 5 times. By round 3 she sounds rehearsed. By round 5 she sounds robotic. The panel then writes "answers felt canned" in the final debrief. She didn't sound canned. She sounded like someone who explained the same 3 bullets to 5 people who claimed to have read them. I'm the recruiter watching this loop every week. I send polished resumes. The panel doesn't open them. The candidate explains herself 5 times. The panel rejects her for "lack of clarity." I rebook the role. Source 12 more candidates. Send 12 more resumes nobody will read. The offer-accept rate keeps falling. Leadership keeps asking me why the pipeline is drying up. It isn't the pipeline. It's the 4 minutes of prep nobody on the panel will spend before they open Zoom.

Sparsh GoyalJun 3, 2026

Hiring is broken. This interviewing system cut 45 day hiring cycles to 3 days. (steal it) Most hiring teams: → Spend 120+ hours interviewing for one role → Get "maybe" answers after 6 rounds → Pay $20K in engineer time per hire → Feels Interviewing = Screening (but its not) AI is solving many productivity problems, but hiring is still at bottom since recruiters feel that basic screening questions using Voice call workflow using VAPI has solved the problem. But here's how I solved it: → Using multi model from Anthropic, creating the system. → Create a role in 2 minutes (not a job description, an actual evaluation criteria) → Build a rubric that tests real skills, not interview performance → How to add 200+ candidates and let AI interview all of them. Any timezone. → What happens when AI goes 5 levels deep on "I scaled to 10M users" instead of saying "cool, next question" → How every rejected candidate gets specific feedback automatically. Not a template. Real reasons. → The exact settings and workflow. Every click documented. This is the system behind: → 95% of candidates eliminated before engineers spend a minute → 40+ days shaved off hiring cycles → 50+ hours of engineering time saved per hire → Every single candidate gets a fair interview and real feedback → 500 applicants. 3 hires. 3 days. <6 hours of human time. Want the full breakdown? (48 hours only) 1. Connect with me 2. Comment "ANTHROPIC" I'll send it. PS — Repost this for priority access