Mickey Graf

@thepanta82 At this point even mixing Claude with codex doesn’t work. Guess human reviewers just go their jobs back

I'm starting to think automated code-review loops are an anti-pattern for the current crop of models. They already have a tendency to overthink things. When you add another layer of overthinking on top, they end up ruining the good work they've done earlier.

@rennyzucker It’s just money

You couldn’t pay me a hundred million to live through this

@ctjlewis Bet drake had a song bout this .. oh yeah say what’s real

Imagine this: Two of your ex girlfriends become best friends and go to Europe together

@jlongster @TotalEnergies should be your n1 customer for this

this uses a graphics programming technique called spherical harmonics. it's physical light it's fun thinking of ways to apply this to brand design & ui. already thinking of a few more interesting ways to apply it x.com/jlongster/stat…

@james406 Rookie numbers tho

"how do you know it's quality code?" is a silly question, as you can see the new code is literally colored green, which means it passes all the tests

@GergelyOrosz Anthropic is really bad at comms so I'd guess it's 4. trying to keep up with OpenAI

This is actually how all of this feels to me. Anthropic either are: 1. untrustworthy for sharing this now not in April when it happened 2. Incompetent for not noticing for 3 months (!!!) 3. Have an agenda for digging out an old incident they never shared before

OpenAI had a damning security incident where their under development AI escaped the sandbox environment and attempted to hack another company (HuggingFace) For some weird reason Anthropic decided to share a similar incident from 3 months ago, only NOW. Something smells off…

This matches where I landed too Last year I worked on the full autonomous loop - from session replay analysis to automatically generating PRDs and handing them off to coding agents. I came away thinking the bottleneck wasn’t the coding. It was the decisions around what to build, whether it was good enough, and when it was safe to ship. That’s what made me stop chasing remove the human from the loopand instead focus on making the small teams (2 eng..) 10x more effective

the real "software factory" isn’t everyone using coding agents. that’s the 2x. the 10x is a different thing entirely. the loop starts closing on its own. 2x = every engineer using agents, ~100% AI-generated code. soon, that’ll be table stakes. you made the same loop faster, but the bottleneck is still human attention at every gate. 10x = the loop self validates. agents stop waiting on you. but it only works if the scaffolding is real: comprehensive CI, evals that actually catch regressions, and end-to-end test environments agents spin up to check their own work. to be clear, 10x is not "fewer humans." it is humans on the decisions that carry real risk and off the rote gates. we still need to hire exceptional engineers, their judgment is the scarce input now, so you spend it where it actually counts. but here is what breaks traditional SaaS instincts: building agents is divergent. fix one thing and 10 new things need evaluating. converging on that locally is a trap. so ship the 80/20 fast, but through controlled early access: opt-in design partners, clear guardrails, tiered rollout. never unfinished work dumped on people who did not sign up for it. then close the loop. feedback-loop agents take that signal, implement it, merge it, validate post-deploy, continuously. and because CI/CD is fast, the gap between "customer said X" and "X is live and validated" collapses. coding at the speed of thought. give agents the same context a human uses to make the call, user feedback, docs, prod monitoring, ephemeral test clusters. encode your review standards as rules the reviewer agent reads. auto-merge the low-risk stuff classified by your reviewer agent. reserve human eyes for billing, auth, the paths where a mistake is expensive. humans do not disappear, they concentrate. incidents stay human-led, business impact's too high to hand off, though agents now dig datadog + telemetry + code and hand you impact + root cause + mitigation in minutes. and someone guards the critical paths: as agentic volume climbs you risk losing deep understanding of your own codebase. the fix isn't to slow down — it is to lock down what's expensive to get wrong. Not all code carries equal risk. at Ramp, this is Inspect, our internal coding agent: deep Ramp context, MCP integrations, CI visibility, and self-validation that boots the UI and tests real workflows. On a recent UI change, it deployed a preview, authenticated as a fixture user, validated desktop and mobile behavior, and attached visual evidence to the PR. the metric I actually care about: fewest human touches per safely merged PR. not lines, not PR count. how much can the loop close on its own without breaking the things that matter and do it super fast? that's the edge.

Daily standups tend to drag the fewer people are present, yet they've become such a standard practice in our industry that few teams dare to question them. If the goal is better coordination in increasingly async teams, shouldn't continuous visibility and transparency reduce the need for a daily status meeting? Maybe the better solution are AI Agents that continuously take care of the visiblity/transparency issue? Is anyone working on this ?

@SamuelBeek That’s why I’m building a solution for this, so people can make sense out of what agents produce

GitHub feels so broken for companies that are CRACKED PR reviews are easily the most annoying and time consuming part of building software right now, I feel like there's so much room for disruption here

I think my copy of Founding Sales expired 9 months ago. Can I still use it?

@eptwts how to win friends and influence people, but I'm more curious about you and what books actually impacted your life?

has any book that actually impacted how you live your life & helped you get to where you are? i'm trying to stack up my bookshelf so gonna do a bulk order of books - any input would be appreciated ideally marketing, business, psychology, etc.

@etnshow @JNeedem co-signed but im nobody on x, only on linkedin :(

JUST IN: @JNeedem has co-signed as our second new alliance member.

@etnshow Guys, if this is a european show then please use CEST

We will be making another announcement at 10:00 BST.

Is „load bearing“ the new em dash or what’s going on with Fable @AnthropicAI?

That’s why I’m bullish on the hybrid human + AI future It’s easy to replace devs with AI agents when the entire business doesn’t rely on the quality of the software product to produce revenue. Once it becomes important you need humans to manage the complexity and context (sic!) in the broader business sense

@etnshow where can I buy some merch?

A lot of mid-market non-tech companies hire engineers to build their own internal software, and I think this is where the move from X devs to 2 devs + agents will happen, not in organizations that ship for millions of users or whose revenue depends on the quality of their software

So many interesting details. Eg: "One thing I’ve heard from some people is “we have two humans and a bunch of agents.” I reply that this isn’t where we’re at: we still have teams with 6-8 people on average." - @katelyn_lesse, Head of Engineering, Claude Platform, Anthropic

@housecor Yes, but humans still need to keep the context of the implementation so they can make further decision

🌶️ Hot take: AI code review > human code review Most devs are too lazy, too busy, or too incompetent to do a good code review. So, the average AI code review is better than the average human review. And it’s getting better every day.

@kimmonismus Remember when it was all over X in January and nobody believed it?