Taylor Eernisse @theirongolddev

I just wish Slack had reliable notifications 😭

GREG ISENBERG @gregisenberg

3 days ago

I wish Slack was: - Agent-first - Beautiful to use - Integrated with agents natively so your Hermes or OpenClaw lives inside it - Huddles worked seamlessly and were fun - Built for teams of 1-3, not just teams of 300 - Truly a second brain similar to Obsidian - Searchable

205 43 941 214K 789

@alexhillman Also, I haven't gotten to pulling that prompt into a skill yet, but clearly I need to move that to the top of the priority list, it's goated

@alexhillman @doodlestein It's a little wild that it's basically just a 6000+ line python script - that is something I will be rearchitecting as the dogfooding activity winds down over the next couple of days

@alexhillman Here's what @doodlestein's /codebase-report skill generated: github.com/theirongolddev… And here's the README: github.com/theirongolddev…

Honestly there’s a ton about the system I don’t understand, I just ran /goal before bed Sunday night with a few sentences based on some thoughts I had and a too-vague completion state 😆 I was mostly just trying out /goal. I continue to be surprised at how thorough the tool is and I’m still working on refining as codex was overly zealous with its redactions 😂 It’s a background project but I’ll eventually get around to generating some architecture diagrams and docs. At a high level it’s a combination of hooks that invoke scripts that analyze input based on a bunch of formal token format lists and mathematical analysis that’s way over my head. A daemon runs in the background to do some of the heavy lifting (this is the part I’m most unclear on at the moment).

@alexhillman Ooooooh this is good

This is a solid breakdown. But what I think people are missing is that since agents are reaching a level of capability similar enough to junior to senior engineers (given the right context and tools), the hard-won lessons around people management, process, units of work, etc. have started applying directly to agentic development. Can one person keep the context of an army of 20 agents in their head at one time? Not effectively. Can one manager keep the context of 20 engineers in their head at one time? Not effectively. Thus far that hasn’t meant the answer was to slow down, avoid parallelization, or make managerial review the bottleneck. So why should the fact that we’re now managing agents instead of humans change the fundamentals of managing work, constraints, and output quality?

Addy Osmani @addyosmani

a week ago

x.com/i/article/2059…

85 362 3K 773K 5K

@GergelyOrosz It’s mind boggling to me that this was greenlit at all Who actually thought this was a genuinely good enough idea to build?

Nobody ever bothers to inspect the session logs of the engineers using AI to generate the bugs Did AI truly engineer the bugs or did the engineer just under specify, pollute context, or yolo the attempt, leading AI to either do exactly as told by context or fill in gaps?

Brandon Carl @brandonjcarl

2 weeks ago

Very interesting data across 2,444 companies. 82% of tokens are spent on AI-generated bugs, rework and review friction.

94 456 3K 636K 1K

All that’s happened is that the giant companies are realizing that the 10-20% productivity gains that the average engineer gains from AI aren’t enough to justify the cost. ClickUp is taking the right approach. Don’t give AI to every engineer equally. Give it to the few who are well rounded enough to use it to generate more value than it costs to use. Fire your lowest performers to free up not just capital but also get the rework they produce off the plates of the rest. The foundations haven’t moved; the basics are still the basics at every level - engineering, business, common sense.

Ricardo @Ric_RTP

2 weeks ago

Microsoft just banned its own engineers from using AI. The tool was literally costing MORE than the humans it was supposed to replace. They lied to you about AI adoption and now the whole narrative is blowing up: Microsoft gave thousands of engineers access to Claude Code six

2K 9K 23K 2.9M 12K

Isn’t this still focusing on the wrong abstraction layer? I don’t particularly care what the code the initial implementing agent wrote looks like, I only care about the code after it’s been through multiple review agents At which point I’ll just look at the PR once an agent has shepherded it through CI to green

James Long @jlongster

2 weeks ago

more and more work is moving into coding agents, I don't live in my editor anymore but you gotta keep an eye on these little goblins, they write bad code. so we built a diff viewer in opencode! available now

74 58 2K 180K 393

@dhe3g The CIA had drones that looked exactly like dragonflies in the 60s

This is easily the best analysis on the topic of AI in the workplace I’ve ever read So many great takes my mind is still reeling

Zeb Evans @DJ_CURFEW

2 weeks ago

Today we reduced headcount by 22%. The business is the strongest it's ever been. So I think it's important to be direct about what I'm seeing and why. First, I made this decision and I own it. I did it because the way to operate at the highest level of productivity is changing,

2K 1K 8K 9.1M 14K

@jeffrey_way @adamwathan Only took two weeks of planning, meetings, and document creation to get here 😂

@jeffrey_way @adamwathan Starting today I will be one-shotting an app for my company over the next eight weeks

Avoid people who make these claims. Avoid companies with execs making these claims. Avoid making these claims! It conflates the output with the tool instead of the one wielding the tool. I can use a set of power tools to create a perfectly square board an order of magnitude faster than a master carpenter can with hand tools. But don’t confuse the squareness of the board with my ability to take that board and build furniture worth passing down to your great-grandkids.

Jason Fried @jasonfried

2 weeks ago

Bragging about how much software you’re shipping with AI is like holding down the shutter button and bragging about how many photos you took.

233 675 6K 242K 577

@regularguyguns @realwitt And this is ignoring the fact that Flock’s cameras are exceptionally insecure, so anyone with a basic understanding of networking can access the cameras, not just police

PLUS it updates independently of the TUI render lifecycle!

The amount of effort I had to use to create a decent statusline in Claude Code compared to the amount needed to create one in pi using GPT 5.5 is wild This took me all of 5 minutes of prompting, and did NOT require me to build an entirely separate go project just to collect the information and render it. I literally vibed it into existence