Online entrepreneur, tech geek, recovering sysadmin, content creator, automation, AI nerd and Lego Bricklink seller.

Joined February 2009
2,692 Photos and videos
Pinned Tweet
If you're in IT or software development in some way, I guarantee you'll learn a thing or two with our 100s of in-depth tech tutorials on adamtheautomator.com.
5
16
159
I have three accounts across Codex and Claude Code. Fable 5 is killing me so I’ve decided a new approach at getting my AI fix.
1
1
285
What's the record for blowing through a Claude Code max subscription's 5-hr window. I did it in 19 minutes with Fable 5 1M Ultracode. Where's my trophy??
1
4
1,061
Spent weeks building a local AI agent to route requests across email, Slack, Codex, and Claude Code. Finally gave up, installed Hermes Agent, added some custom routing, and it just worked. Sometimes the smartest engineering decision is to stop engineering.
2
1
5
566
Codex and Claude Code both have /goal now. But /goal only works well when success is deterministic. One useful pattern: create a subagent that evaluates the result and returns a score. Then your goal becomes simple: “Keep going until the evaluator returns 8/10 or higher.”
262
Mark my words. One day we won’t have software documentation. Instead, vendors will all provide agent skills so your agents can get up to speed fast.
1
3
227
Judging an employee's productivity on the number of tokens they consume from AI models is exactly the same as judging their productivity on lines of code written.
1
3
435
Outsmart Attackers: A Guide to Advanced Ransomware Protection with Azure NetApp Files Learn how to configure advanced ransomware protection in Azure NetApp Files using ML-based detection to catch... adamtheautomator.com/outsmar…
1
260
Quick #Codex harness hack: Set a /goal for GPT5.5-low for some big feature. It may not finish because it's on low thinking but it will iterate through LOTS of ideas fast. After awhile, stop it, change thinking to Extra High and summarize the convo. You'll get tons of insight.
1
1
331
I run my life with AI agents, which meant building tools to let them interact with the world. I’ve standardized them by sharing auth, caching, browser automation and more features. Excited to finally make this repo public for everyone! 👇github.com/adbertram/cli-too… #Codex #ClaudeCode
5
1
4
535
The irony in this image is unmistakable.
4
484
Building AI agents? Test your instructions on a smaller model like GPT-mini or Haiku. Frontier models hide sloppy prompts; smaller models expose every unclear step fast because they'll fall down immediately.
2
304
If you've purchased #PowerShell for Sysadmins, I'd truly appreciate it if you could submit a review on Amazon. Thank you! amazon.com/review/create-rev…

278
I honestly can think of zero reasons why AI agents need anything but skills, cli tools and a model. Everything else seems redundant to me. Change my mind.
11
13
785
As soon as I think we’ve seen the peak of what we can slap an AI label on, I see this. I totally need AI in my dryer.
1
1
3
553
I'm starting to assign a personality to my Codex app. I asked it/him? to create himself a Slack profile photo and set it and this is what he came up with. Introducing, Cody, my assistant!
1
245
Adam Bertram retweeted
🚨 OpenAI 's own engineers just showed how to actually use OpenAI Codex properly. 60 minutes. free. built by the people who contribute to made it. watch the masterclass. bookmark it. worth more than every $900 coding course you almost bought. you’ve been using Codex like a simple coding tool… while it’s actually a full software engineering system. watch this, it could the best 62 minutes of your life:
53
350
2,802
291,335
After using Claude Code as my daily driver for nearly a full year, I’ve switched to the Codex app. Partly because GPT-5.5 is phenomenal, partly because of the Codex app, and mostly because of usage limits. You can make your dollar go so much farther with Codex.
22
5
337
9,889
It is the end of an era. I only thought it was fitting that Codex kill it's competitor.
3
139
Me: Claude Code, make me a sandwich. Claude: You're right to want one. Bologna feels right because... training data. Done. Me: Codex, make me a sandwich. Codex: First, let's audit the cupboard, compare breads, inspect fillings, define requirements, and draft a sandwich plan.
3
367