Mihail Eric

@mihail_eric

Head of AI @monacoGTM, @stanford teaching AI productivity for 32K+ devs themodernsoftware.dev, YC S24, @ConfettiAI (acq'd), ML @amazon @stanfordnlp

San Francisco, CA

mihaileric.com

Joined September 2016

Replies Reposts Media Articles

Pinned
Mihail Eric
@mihail_eric
Sep 2
I’m excited to finally announce the newest edition my Stanford course 𝗧𝗵𝗲 𝗠𝗼𝗱𝗲𝗿𝗻 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿. It has been 9 months in the making. Last November, with the release of Claude Opus 4.5, coding agents experienced a step function improvement in
240
Mihail Eric
@mihail_eric
20h
One of my favorite new interview questions: ask someone to write the prompt for an engineering problem that's impossible to one-shot correctly. In a world where writing lines of code is commoditized, clear communication is the new alpha. Prompting is a medium where the prompter
3
Mihail Eric
@mihail_eric
Sep 4
𝘈𝘐 𝘥𝘦𝘴𝘱𝘦𝘳𝘢𝘵𝘪𝘰𝘯 is what we call it when coding agents, in an attempt to please, put forward plausible and hopelessly wrong theories for bugs. Sometimes you have to just roll up your sleeves the old-fashioned way and look at the code.
Monaco
@MonacoGTM
Sep 4
An expensive phantom script that nobody ever runs. A job that has a 1% chance of hanging forever. When AI struggles with hard problems, it gets desperate and tries to solve you rather than the problem. This post shares how we fought AI desperation while debugging two of the
2
Mihail Eric
@mihail_eric
Sep 4
This is one of the most sophisticated cost breakdowns for measuring agentic coding impact and ROI in an enterprise I've seen. @UberEng published how it runs its software factory. Over 70% of its pull requests now come from agents, usage grew 7x since February, and total spend
6
Mihail Eric
@mihail_eric
Aug 31
Code is no longer the bottleneck. Defining intent and reviewing output are the new constraints. Anthropic's Applied AI team published a stage-by-stage playbook for the AI-native SDLC. Every stage ends by committing an artifact the next one reads: 𝐏𝐥𝐚𝐧: Whoever has the idea
5