What AI Tools Are Engineers Using?

Every few months someone publishes a new roundup of the best AI tools for developers. They all list the same tools, describe the same features, and skip the parts that actually help you make a decision. This is a different take: what engineers are actually using, what the real tradeoffs are, and a curated list of resources worth your time for staying sharp and interviewing well.

How Have AI Tools Changed?

Things shifted a lot in the past eighteen months. A 2026 Pragmatic Engineer survey of nearly a thousand engineers found that Claude Code went from essentially unknown to the most-used agentic coding tool among their readership in under a year. Cursor crossed $2B ARR. GitHub Copilot, which was basically synonymous with “AI coding tools” two years ago, is now one of several serious options rather than the automatic default.

The breakdown by company size is fairly predictable: large enterprises (10,000+ employees) still lean toward Copilot, mostly because enterprise procurement already approved it. SOC 2, audit logs, IP indemnification, SSO. Startups and smaller teams are much more likely to be on Cursor or Claude Code, where individual engineers have more influence over what they use.

Here’s how the three main tools differ in practice:

GitHub Copilot is the one most engineers have already used. Solid autocomplete, mature IDE integration, and if you’re joining a company with established security requirements, it’s probably already been approved. The honest critique is that Copilot’s core experience is no longer best-in-class for complex tasks. Both Cursor and Claude Code handle larger context windows better. It’s the safest enterprise choice, not the most powerful developer choice.

Cursor is an AI-native IDE built on VS Code. When a whole team is on it, the experience is genuinely impressive. Agent mode handles multi-file edits well, and it supports multiple underlying models including Claude Opus 4. The friction is that it asks everyone to standardize on one IDE, which is harder to sell than installing a plugin. Team tier pricing is $40/seat/month.

Claude Code runs in the terminal as an agentic tool rather than inside an IDE. It reasons well across large codebases and is currently the highest-satisfaction tool among engineers who use it regularly. The Pragmatic Engineer survey found 46% of users love it, compared to 19% for Cursor and 9% for Copilot. It’s particularly good for tasks that span many files: large refactors, migrations, deep codebase audits. It’s not the right fit for junior engineers who need guardrails. It works best when you know enough to verify what it’s producing.

The teams getting the most out of AI tooling tend to use multiple tools with some intention behind the choice. Copilot or Cursor for daily completions, Claude Code for heavier architectural work. Teams on one tool for everything have usually had the decision made centrally.

What This Means If You’re Interviewing

A few things worth paying attention to when you’re talking to companies:

If a team hasn’t thought about AI tooling at all, ask why. Some startups haven’t made a decision yet because they’re moving fast, which is fine. But if you’re talking to a twenty-person engineering team and nobody has opinions about this, that’s worth noting.

If they’re on Copilot exclusively at a smaller company, it might be legitimate security requirements. Or it might be low engineer autonomy over tooling decisions. Both are worth understanding before you join.

If they mention Cursor or Claude Code, follow up and ask what they actually use it for. Engineers who have real answers to that question have been thinking about their workflow. That usually carries over into everything else they do.

The Resources Worth Your Time

A lot of noise exists in the technical interview prep space. Here’s what’s actually useful, organized by what you’re trying to do.

Algorithms and coding interviews

**Neetcode.io **— Cleaner than open-ended LeetCode grinding. The roadmap is well-structured, the video explanations are good, and the problem groupings by pattern (sliding window, two pointers, trees) build intuition better than random mediums.

**Tech Interview Handbook** — Free, well-maintained, and the most honest high-level guide to the full interview process. Covers what to study, how to study it, and what the different stages of a loop are actually evaluating. Not a substitute for practice, but a useful orientation before you start.

**Blind 75 / Neetcode 150 **— Specific curated problem lists, not platforms. The Blind 75 is the original high-signal set. Neetcode 150 extends it with better categorization. If your time is limited, these are a better bet than open-ended grinding.

System design

**ByteByteGo** — Alex Xu’s materials (book, newsletter, YouTube channel) are the current standard for mid-to-senior prep. The YouTube channel covers a lot for free. If you’re interviewing for senior or staff roles and haven’t touched system design prep recently, start here.

**interviewing.io **— Paid, but the best mock interview resource available. You can book sessions with actual engineers from major companies, anonymously. If you’re close to ready and just need real feedback to calibrate, it’s worth it.

**Exponent** — Good for engineers moving into staff or principal territory, or companies where system design is heavier than the coding round.

Staying current on engineering practice

**The Pragmatic Engineer **— The most reliable source for honest, data-driven coverage of what engineering actually looks like at scale. Their AI tooling surveys are referenced throughout this piece because they survey real engineers and publish actual numbers.

**Lenny’s Newsletter **— Worth following if you work at or are targeting product-focused startups. Less purely engineering, but useful for understanding the product context around your work.

**A Life Engineered **— Practical, honest, from someone who’s actually been in these rooms. Especially useful for understanding what interviewers are evaluating rather than just what they’re asking.

Understanding what companies actually look for

**Levels.fyi **— Obviously useful for comp benchmarking. Also useful for understanding leveling criteria across companies, which tells you a lot about what “senior engineer” means at a specific place.

**Glassdoor interview reviews** — Underused. The specific interview question logs for a company, sorted by recent, give you a real-time read on what the loop looks like and what tripped other candidates up.

**Blind **— Low signal-to-noise ratio, but the signal when it exists is high. Compensation transparency and honest company takes from people currently working there.

One Thing About AI-Assisted Prep

Using AI to help you prepare is fine and probably useful. Using AI to simulate a mock interview where it tells you your answer was great is not. The tools that help are the ones that push back, either a real person through interviewing.io or a system actually designed to critique. If you’re using Claude or GPT to understand why a solution works, that’s useful. If you’re using it to rehearse while it tells you everything sounds good, you’re practicing the wrong thing.

Most engineers who are “preparing” are either grinding LeetCode without structure or reading generic advice that won’t change anything. The list above is a shorter path to where you’re actually trying to get. If you’re actively looking and want to talk through which roles in the current market actually fit what you’re after, that’s what Fonzi does.