iiinova — where your digital self keeps evolving

Robert Herbig @robertherbig.bsky.social · 23/09/2026

A company decided developers just translate specs to code, so they had AI write the specs first. All of them. Five months later: a million lines of markdown, no code, and a spec directing payroll data into the browser cache. Beta's due Nov 1. My prediction is in writing: sep.com/blog/a-milli...

sep.com

A Million Lines of Markdown and No Code

A real company spent five months generating a million lines of AI-written specs, and no code, betting a November deadline on spec-driven development. What their story reveals about waterfall, iteratio...

010

Robert Herbig @robertherbig.bsky.social · 25/08/2026

The "AI killed junior devs" discourse keeps telling juniors to level up or be replaced. Wrong target. Turning juniors into seniors is the firm's job, and our data shows what happens when a firm actually does it: 95% of new grads who stay 6+ years reach senior or above. sep.com/blog/everybo...

sep.com

Everybody Agrees You Should Hire Junior Engineers. Almost Nobody Does.

SEP has hired six to nine junior engineers every year since 2020. Here's the actual math behind why that still makes sense with AI in the picture.

031

Robert Herbig @robertherbig.bsky.social · 02/07/2026

AI changes the balance between writing code and reviewing it. Writing code got faster. But did shipping features get faster? The solution isn't just using AI more or harder, it's restructuring around where humans are still the bottleneck. sep.com/blog/faster-...

sep.com

Faster Coding Means Slower Everything Else

AI coding productivity bottlenecks are shifting the balance from writing code to reviewing it among developers.

100

Robert Herbig @robertherbig.bsky.social · 10/06/2026

"Do we need custom AI or off-the-shelf?" is the wrong question. The right one: which parts of this problem need custom work? Because the answer is usually different for each piece. Wrote up how to tell the difference: sep.com/blog/custom-...

sep.com

Custom AI vs. Off-the-Shelf: Stop Asking If, Start Asking Where

They were sure they needed custom AI. They were wrong. How to tell which parts of a system to build, and which to buy off-the-shelf now.

000

Robert Herbig @robertherbig.bsky.social · 27/05/2026

New post: We're Picking Tools for a Different Developer Now When agents write most of the code, tool selection criteria shift in ways most teams aren't accounting for yet. I talked to a bunch of practitioners about what actually makes a tool agent-friendly. sep.com/blog/were-pi...

sep.com

We're Picking Tools for a Different Developer Now

Read We're Picking Tools for a Different Developer Now to learn about AI & Data, Programming.

431

Reposted by Robert Herbig

Charles @tybar.bsky.social · 14/05/2026

I've been thinking about how new junior software engineers keep learning fundamentals as they work with agentic AI tools. I haven't found any good answers yet, but talking to respected folks in the AI community (like @robertherbig.bsky.social ) has been interesting. Thoughts?

011

Robert Herbig @robertherbig.bsky.social · 05/05/2026

The AI agent's brain gets wiped every session. I realized I needed to make the starting point better. Asked a bunch of engineers how they handle this. Wrote up the spectrum of approaches + where it breaks down. sep.com/blog/treat-y...

sep.com

Treat Your AI Workflow Like You Treat Your Code

Learn effective AI Workflow techniques to avoid errors and improve communication between you and your AI co-worker.

000

Robert Herbig @robertherbig.bsky.social · 20/04/2026

Wrote about something I've been thinking about a lot: a team that went all-in on shared AI infrastructure found it significantly increased human collaboration. Other teams using the same AI tools, but individually finding their own way? No change, or even less collaboration. sep.com/blog/the-ai-...

sep.com

The AI Collaboration Paradox

Does AI make teams more or less collaborative? It depends on how they adopt it. SEP's AI practice lead breaks down the pattern across real engineering teams.

010

Robert Herbig @robertherbig.bsky.social · 02/04/2026

@steve-yegge.bsky.social 's AI Vampire article resonated with a lot of people. The energy drain is real. But I don't think the explanation most people landed on is right. I asked teams deep into agentic workflows what's actually causing the fatigue: sep.com/blog/where-a...

sep.com

Where AI Fatigue Actually Comes From

Read Where AI Fatigue Actually Comes From to learn about AI & Data.

000

Robert Herbig @robertherbig.bsky.social · 12/02/2026

According to Microsoft, GitHub, and the DORA Report, actual coding is only 8% of the software delivery cycle. Most teams' AI strategy is optimizing for that 8%. On March 4 we're sharing real examples of what happens when you go after the other 92%. Free webinar, 11 AM EST: sep.com/events/beyon...

sep.com

Beyond Code Gen: Rethinking AI's Role in Software Development

000

Reposted by Robert Herbig

Pigmy Wurm @pigmywurm.bsky.social · 03/04/2025

So this chart popped into my head as a very rough tool. Not perfect and I'm not sure on the exact breakpoints for different Brackets (especially 4 & 5), but I feel it's a decent visualization.

Color coded chart showing the rough stage of the game for each bracket.
The colors are labeled:
Dark Green-Safe: No one is winning or making must answer plays.
Light Green-Developing: No one is going to win, but high impact plays can generate a lot of advantage.
Yellow-Caution: Winning isn't likely but it's possible and the game is moving in that direction.
Orange-Danger: Winning is getting more likely, especially if a player is left unchecked.
Red-The End is Nigh: a Player winning the game is expected.
Dark Red-Why aren't you Dead!?: Most of the time the game is supposed to be over by this point.

Bracket 1: Safe turns 1-3, Developing 4-7, Caution 8-9, Danger 10-11, End is Nigh 12-15.
Bracket 2: Safe turns 1-2, Developing 3-6, Caution 7-8, Danger 9-10, End is Nigh 11-13, WAYD? 14+
Bracket 3: Safe turn 1, Developing 2-3, Caution 4-6, Danger, 7-8, End is Nigh 9-12, WAYD? 13+
Bracket 4 Developing turn 1, Caution 2-3, Danger 4-5, End is Nigh 6-8, WAYD? 9+
Bracket 5 Caution turn 1, Danger turn 2, End is Nigh 3-5, WAYD? 6+

69113

Reposted by Robert Herbig

JA Westenberg @joanwestenberg.com · 23/10/2025

Absolutely obsessed with framing your employees sleeping on the floor as a humble brag “Look how hard my team works” no dude you’re just bad at project management. Competent companies ship products without cosplaying being homeless

A tweet by Mark Kretschmann showing two photos of the xAI office on Grokipedia release day. The office is filled with small camping tents set up between desks and along hallways, suggesting that engineers have been sleeping there while working late. The caption notes it’s early morning and the engineers are still in their tents, joking that a good night’s sleep does wonders.

41880781798

Robert Herbig @robertherbig.bsky.social · 08/10/2025

I'm super excited to announce I'll be speaking at @codemash.bsky.social on January 13-16, 2026! My talks are "Escaping the Trap of Self-Sabotaging Meetings" and "We Shipped an AI Agent. Users Didn’t Care." (events.codemash.org/2026CodeMash...). I hope to see you there! #CodeMash #CodeMash2026

021

Reposted by Robert Herbig

Erin Reed @erininthemorning.com · 01/10/2025

Honestly “Republicans are shutting down the government to hide the Epstein files” is pretty good messaging.

2532591023

Robert Herbig @robertherbig.bsky.social · 10/07/2025

I'm very excited to be at the dev up 2025 conference in St. Louis in less than a month (August 6-8)! I'll be giving two talks about AI: "What Are AI Agents, Anyway?" and "When is a Regular Expression Better Than Artificial Intelligence?". Get your ticket here: www.eventbrite.com/e/dev-up-202...

010

Reposted by Robert Herbig

Ed Zitron @edzitron.com · 11/06/2025

This lawsuit is brutal. One of the exhibits is the letter Universal sent to Midjourney, including a quote from a 2022 Forbes interview that I bet midjourney regrets

Universal did not permit, authorize, or license Midjourney to copy or use any of Universal's copyrighted works in connection with any version of its Image Service or to otherwise exploit Universal's works at any time. You acknowledged as much in a September 16, 2022 Forbes interview in which you confirmed that Midjourney never sought any copyright holders' consent to use or exploit their works:
Did you seek consent from living artists or work still under copyright?
No. There isn't really a way to get a hundred million images and know where they're coming from. It would be cool if images had metadata embedded in them about the copyright owner or something. But that's not a thing; there's not a registry. There's no way to find a picture on the Internet, and then automatically trace it to an owner and then have any way of doing anything to authenticate it.
See Rob Salkowitz. Midiourney Founder David Holz

220111314018

Robert Herbig @robertherbig.bsky.social · 02/06/2025

I'm super excited to announce that my new APIsec University course "Building‍ Security into AI" is now live! www.apisecuniversity.com/courses/buil... You check out a brief overview here: www.youtube.com/watch?v=1wZI...

apisecuniversity.com

Building Security into AI | APIsec University

Learn how to build security into AI from the ground up. This course covers AI vs. traditional app development, threat modeling for AI systems, real-world breaches, and essential cybersecurity practice...

011

Robert Herbig @robertherbig.bsky.social · 30/05/2025

In case you missed it, @stirtrek.com is an amazing conference and they recorded it! You can watch E. Nigma talk about Uncertainty: www.youtube.com/watch?v=8IMV...

youtube.com

Can We Learn to Manage Uncertainty? Probably! - Stir Trek 2025

YouTube video by Stir Trek

041

Robert Herbig @robertherbig.bsky.social · 19/05/2025

There are just two short days until APISecU's APISEC|CON (conf.apisecuniversity.com). I'll be introducing my new course on AI Security. I hope to see you there!

000

Robert Herbig @robertherbig.bsky.social · 02/05/2025

I'm very excited to speak about "Can We Learn to Manage Uncertainty? Probably!" @stirtrek.com 2025! I'll be in the Captain Marvel room at 2:00pm. The slides are available here: www.dropbox.com/scl/fi/90vfx... I hope to see you there!

dropbox.com

042

Robert Herbig @robertherbig.bsky.social · 23/04/2025

Vibe coding feels like peak Agile: working software, fast feedback, no ceremony. Just a dev and their AI, iterating in real time. But beneath the speed, the structure breaks. Why vibe coding isn’t Agile, but just frAgile: sep.com/blog/vibe-co...

sep.com

Vibe Coding Is the Purest Form of Agile (And That’s the Problem)

Check it out! Read "Vibe Coding Is the Purest Form of Agile (And That’s the Problem)" to learn more about Artificial Intelligence at SEP.

010

Robert Herbig @robertherbig.bsky.social · 03/04/2025

Vibe coding is fast - super fast - but it breaks down super fast, too. I wrote about why AI-generated code becomes unworkable over time, and how to keep things stable, extensible, and fast (I call it Signal Coding!). Why Vibe Coding Fails - and How Signal Coding Fixes It: sep.com/blog/vibe-co...

sep.com

Why Vibe Coding Fails - and How Signal Coding Fixes It

Check it out! Read "Why Vibe Coding Fails - and How Signal Coding Fixes It" to learn more about Artificial Intelligence at SEP.

101

Reposted by Robert Herbig

Jordan Thayer @jordanthayer.com · 10/02/2025

Everyone in my field could benefit from reading this, but especially this excerpt: "In fact when people give accounts of sexist and racist harassment they are often dismissed as having a wrong or faulty perception, as not receiving the intentions or actions of others fairly or properly."

021

Reposted by Robert Herbig

Gregory Travis @gregorytravis.com · 18/01/2025

IMPORTANT The US CDC has announced that, going forward, reported SARS-CoV-2 wastewater levels will be normalized to an endemic baseline "Zero" on this baseline will be levels in the previous year @evanblake17.bsky.social @benjaminmateus.bsky.social

59774445

Robert Herbig @robertherbig.bsky.social · 17/01/2025

Conference speaker friends! I am looking for Illinois-based software conferences to speak at. If you have any you would recommend, I'd love to hear about them. Thank you in advance!

010

Robert Herbig @robertherbig.bsky.social · 16/01/2025

I'm excited to learn more about data structures (that's a sentence I never thought I'd type!) with @jordanthayer.com at @codemash.bsky.social 2025 #codemash2025 #dothemash

The speaker, Jordan Thayer, introducing the session topic of data structures to the audience

000

Reposted by Robert Herbig

Marisa Kabas @marisakabas.bsky.social · 05/04/2024

NEW: I've republished the Kansas Reflector column still being blocked on all Meta properties. It's a piece critical of Facebook's ad policy on climate change content that triggered a site-wide ban yesterday. Read/share widely: www.thehandbasket.co/p/kansas-ref...

thehandbasket.co

Here's the column Meta doesn't want you to see

Read the Kansas Reflector's piece about Facebook's climate change ad policy

25476292

Reposted by Robert Herbig

Phil Foglio @thephilfoglio.bsky.social · 01/01/2024

#Webcomics Hey kids! Here's a fun game! They just dropped a list of All the artists that Midjourney admits to having scraped to train its A.I. engine. Are YOU on it? (We are) Do you have a good lawyer? (We do) (storage.courtlistener.com/recap/gov.us...)

storage.courtlistener.com

14527351970