Join me on this AI journey. As I learn the basics and move into more complex topics, feel free to follow along!
North Carolina
Joined July 2011
- Tweets472
- Following395
- Followers173
- Likes492
Pinned Tweet
Life update: I just got promoted to Applied AI Engineer.
I’ve always loved the work I do. Over the past year or so I’ve been going deeper into agentic systems, prompting, vibe coding, and AI engineering, and this new role is the first time it feels less like a job I love and more like something I’d do anyway.
Pretty grateful for that.
#AI #AgenticAI #AIEngineering #AppliedAI #CareerGrowth
Chris retweeted
Cribl Detect brings detection posture management, #AI-powered triage and investigation, SOAR, and compliance capabilities together on the Cribl platform.
Help analysts move from alert to answer faster, without manually correlating data across tools: bit.ly/4rWVjbG
Do you think loop and graph engineering are transitioning to software factory engineering? Seems like that's where everybody's going at this point, and honestly, I'm struggling with it.
Transitioning from @resend to Boxox today! Works with all the technology that I already use! I recommend giving it a try yourself, especially is you already use Cloudflare!
Released boxox.dev today.
Let's you have api access to cloudflare's email services to use for whatever you want. Has a mail client, agent email, cli, mcp, all the letters.
It's free unless you want to create inboxes, then it's @Cloudflare 's deserved 5 bucks.
I've been using Dots since release, but haven't thrown a ton of "real" work at it yet.
So far, we've built a family meal planner with a cookbook and grocery list, connected it to Dots, and planned meals around what my family will actually eat and what we already had.
Then it put together a Walmart order from that plan, I approved checkout, and the groceries showed up at my door.
There were a few gaps. The planned meals weren't initially saved in the cookbook, so we fixed that and reviewed the UX for other rough edges.
Still some directing and checking involved, but going from "help me figure out dinner" to a working app and groceries at the door is a pretty solid start.
Today I called my @OpenAI Dot while I was driving and just had a 20-minute conversation about some ideas I was having on projects that I had been working on with GPT. It was super beneficial and I thought we had made some progress on some items as well as flagged some future brainstorming ideas. I ended the conversation with "follow up with me through the phone app" and figured I would hear from it later. I heard nothing, and when I followed up 6 hours later, it was unable to tell me what I had thought about, what I had come up with, and the ideas that I had thrown at it. Slightly disappointed from the lost flow. I still have faith in Dot, but hopefully it gets better than this.
We should direct our AI, not let it direct us. Before accepting its output, check it against your original intent. Write that intent down so you can spot drift. Keep the design your own. Let AI challenge your thinking, but make the decisions yourself.
#AI #HumanInTheLoop
I want Google's new model Argon to be mind blowing. I'm very deep into the Google ecosystem and I want to believe in them. The time between announcement and general availability have made me concerned, to be honest. I know we're in a fast moving space when it comes to model releases, so maybe it's just me.
#google #argon #Gemini
These are really solid in my opinion and has the potential to consolidate/relieve a few separate tools.
Roundup of Day 2/
2.1/ Approve for me (auto-review) is now included and does not use usage. Can be between 2-10% of plan when used. Also better for you.
2.2/ Simplified API for builders.
2.3/ Meeting notes integrated.
2.4/ Decisions API live for builders. Will use in the app to improve the experience.
This is the type of feedback that is useful, not "my COD clone doesn't look as pretty with GPT 6.1 Sol", actual independent feedback with metered results.
we found opus 5.5 to be 2X as efficient as gpt-6-sol in our internal cfo.ai evals so i was not expecting 6.1-sol to improve on that much
but to see it 3X the efficiency of opus 5.5 and 2.5X against even sonnet 5.5 is actually 🤯
this is an insane point release
It's nice to see the GPT 6.1 Sol doomers calming down after actually using the model and seeing the cost savings.
So, I make a deal of watching "vibe coding" streams. It's great background noise while I work, and some of the things built are just cool. What's frustrating is the lack of human design from what they're building. Too often it's prompt, accept, repeat. The agent ends up steering the design instead of the human. We should lead with a clear outcome and direct the tool. Not let the tool decide the destination.
How does CloudAtlas do with token usage vs discounts? I've tried a few "discounted" token routers that somehow ended up being more expensive against something like OrcaRouter/OpenRouter.
