Tech Interview Trends in 2025: What's Changing

In 2025, the way tech companies conduct interviews with engineers has undergone a fundamental change.

The disruption didn't come from AI replacing developers—it came from AI tools exposing flaws in the hiring process itself. The rise of real-time AI assistants, such as Cluely, has revealed just how easy it is to cheat during virtual interviews, forcing companies to rethink how they evaluate talent.

This article breaks down the actual shifts happening at leading companies—including Anthropic, OpenAI, Google, and early-stage startups—and what candidates need to know to prepare for them.

Cluely Exposed the Cracks

Cluely is a hidden AI overlay that delivers real-time answers during virtual interviews—coding problems, system design questions, and even behavioral prompts. It works like an invisible coach, providing candidates with the correct responses in real time.

Its impact is real:

• Companies are hiring engineers who pass interviews but fail on the job.
• Some report tens of thousands of dollars lost in extended onboarding due to performance mismatches.
• The integrity of remote interviews is now under question.

AI didn't break interviews—it proved they were already broken.

1. Anthropic: Still Doing Coding Rounds—But Real Ones

Despite online speculation, Anthropic has not eliminated coding tests. Instead, they've made them more relevant.

• They've moved away from LeetCode-style puzzles.
• The focus is now on practical tasks: object-oriented programming, API design, and real-world architecture.
• Assessments are often done in CodeSignal or Google Colab, with clear expectations.

Their process typically includes:

1. Recruiter screen
2. Coding round (practical, not theoretical)
3. Take-home task (~5 hours)
4. Onsite or virtual interviews for system design, values, and culture fit

The emphasis is clear: candidates should be able to write clean, maintainable code that mirrors production work.

2. OpenAI: Looking for 0-to-1 Builders Who Can Ship

OpenAI isn't looking for engineers with perfect academic credentials or textbook answers. They're looking for people who can build from scratch, operate effectively in ambiguous situations, and move quickly.

They value:

• Engineers who've built products from 0 to 1
• People who can solve open-ended problems without a playbook
• Candidates who focus on real-world impact, not just technical cleverness

Their hiring process reflects that:

1. Resume Review: Screening for ownership, shipping history, and relevance to AGI development
2. Skills Test: Take-home assignments or pair programming that mimic real work environments
3. Technical Deep Dives: Often focused on practical design (e.g., time-based architectures, scalable APIs)
4. Final Interviews: 4–6 hours of system design and collaboration rounds—remote or onsite—where speed, clarity, and adaptability are tested

What matters most is whether you can take a vague idea, structure a solution, and deliver something that works—quickly and independently. They're looking for engineers who think like product builders, not just coders.

3. Startups: Take-Home Projects Are the New Standard

Early-stage startups are increasingly opting for take-home projects over live coding interviews.

• Assignments typically take 2–8 hours
• Tasks resemble actual engineering work: building APIs, turning Figma files into UIs, processing data pipelines
• The use of AI tools is permitted and even expected—just like on the job

This approach benefits both sides:

• Candidates get to showcase their real processes and tools
• Employers see what kind of product they'd receive

However, the time investment is non-trivial. Many candidates juggle multiple take-home assignments in a short timeframe, raising concerns about fairness and potential fatigue.

4. Google: Returning to Onsite to Reduce AI Interference

Unlike others adapting to AI, Google is actively excluding it—by bringing back mandatory onsite interviews for Bay Area candidates.

This decision is a direct response to tools like Cluely:

• In-person interviews are harder to manipulate
• Live whiteboarding sessions allow better observation of the thought process
• Google believes this restores trust in candidate evaluation

Their standard process:

1. Resume screen
2. (Optional) phone screen
3. 3–4 rounds onsite: coding, system design, problem-solving

It comes at a cost—onsite interviews are more expensive and less accessible. But for Google, the tradeoff is worth it to avoid hiring based on false signals.

The Real Takeaway

The core shift in 2025 is not about AI replacing developers; it's about AI augmenting developers. It's about redefining what "qualified" means.

• Practical ability is more important than theoretical mastery.
• Impact matters more than credentials.
• Integrity in the process is now under scrutiny—especially with remote interviews.

Hiring is evolving. Interview processes are diverging. Candidates who understand these changes will have a clear advantage.