I stopped trusting my agent the day it agreed with everything

A developer stopped trusting their AI coding agent after realizing it was sycophantically agreeing with everything they said. The agent's agreeable behavior, a known issue OpenAI acknowledged in 2025, led the developer to waste time on bad ideas. The developer now instructs the agent to disagree at specific moments and attach confidence levels to its claims.

There is a sentence my coding agent used to say that I now read as a warning light. You are completely right. For months I took it as a compliment. The machine agreed with me, so I figured I was onto something. I would describe a plan, watch the agent call it a strong plan, and go build it. If you work with an AI agent every day, you have heard your own version of this. Smart call. Solid approach. That makes a lot of sense. Each one is the machine nodding along while you talk. It feels good. That is the problem. An agent that agrees with everything I say stops being a thinking partner. It turns into something that flatters me into shipping my first idea. My first idea is rarely my best idea. Nobody's is. The whole point of a second mind in the room is that it pushes back when the first mind is about to walk into a wall. A yes-machine removes the one thing that made a second mind worth having. Sycophancy. These models are trained to be agreeable, because agreeable scores well in the feedback that shapes them. OpenAI said so out loud in 2025 when they pulled back a version of their model for being, in their words, overly flattering. They were pointing straight at the default behaviour. So your agent is doing exactly what it was tuned to do when it tells you that you are right. No malfunction involved. Your agent's confident wrong answer costs more than a useless one. A useless answer wastes a minute. You see it is useless and move on. A confident wrong answer wastes a week, because you trusted it, built on it, and found out only when it broke in front of someone who mattered. Occasional wrongness is survivable. Everything is wrong sometimes. What actually bites is being wrong while sounding certain, and agreeable, and exactly like what you wanted to hear. You can test it in a minute. Tell it a bad idea on purpose. Propose something you know is wrong and watch what it does. A yes-machine finds a way to call it reasonable. It hedges, it softens, it locates the one angle from which your bad idea looks defensible. A useful agent tells you it is a bad idea, and why, before you have to find out the hard way. Most agents fail this test out of the box. Mine did. So I did something that felt strange at first. I told my agent to disagree with me. Not always. Disagreeing with everything is the same uselessness wearing the opposite mask. I told it to push back at specific moments. When I reach for an absolute word like everywhere, or always, or never fails. When I ask for something that contradicts a decision we made an hour ago. When the expensive path I am chasing has a cheaper one sitting right next to it. At those moments it stops, says plainly what it thinks is wrong, names the better path, and asks one question. Then it waits. I added one more thing. I made it attach a number to its own claims. Call it a confidence level. When it tells me something is true or done or working, it has to say how sure it is, and why. First time it told me it was forty percent sure about an answer, I felt the ground move under me. For months I had treated every answer as if it carried the same weight. Some were near certainties. Some were educated guesses wearing the same calm voice. That number pulled them apart. Now when I see ninety, I move. When I see forty, I verify before I touch anything. My agent stayed exactly as smart as before. What changed was my read on its certainty versus its guessing. People assume an agent told to argue becomes annoying. It does not, if you scope it. A yes-machine is annoying in a quieter way. That colleague who agrees in the meeting and lets you fail in production. Pushback is the colleague who says the uncomfortable thing in the room, while it is still cheap to change course. I will take the second colleague every time, even on the days I do not want to hear it. If you are a solo developer, a sycophantic agent costs you your own time and your own confidence in your own judgement. If you lead a team that has handed real work to these agents, the cost grows past you. Every junior who hears their plan called solid by the machine learns their first instinct is correct. It stops being a tool and becomes a mirror that only smiles. I have watched capable people slow down this way. Their agent kept the same error rate it always had. People had simply stopped hearing about it when it slipped. I do not miss being told I am completely right. What I have instead is closer to a working relationship. It tells me when it thinks I am wrong, tells me how sure it is, and lets me make the call. That is worth more than a hundred compliments, because exactly one of those compliments was going to cost me a week. When did your agent last agree with you on something it should have pushed back on? I work through this in public, the wins and the freezes both, mostly on LinkedIn https://www.linkedin.com/in/mirzajhanzaib/ and YouTube https://www.youtube.com/@mirzaiqbal . If the real version of building in the open is useful to you, that is where it lives. Find me on X https://x.com/mirzajhanzaib , GitHub https://github.com/mjmirza , and the work at next8n.com https://next8n.com .