Sign in

Dave Davies

@onlineinference.com
472 followers 137 following 340 posts

I'm an SEO who gets to work with people who teach machines how to think.

PostsRepliesMedia
Dave Davies @onlineinference.com · 04/05/2026
“The system doesn’t fail because it stops; it fails because it doesn’t know when to stop.” Many agent failures aren’t crashes 🤕; they’re loops➿. If you’re building agents: 🔄 max retries 💸cost ceilings ⏪fallback paths Or you scale failure. @wandb article in the comments👇
100
Dave Davies @onlineinference.com · 12/02/2026
As someone who vibe codes … this is great advice from @weightsbiases.bsky.social . Prompting isn’t the bottleneck 🍾. Context is. RPI (🧪Research → 🤔Plan → 🛠️Implement) + 🤏intentional compaction = ➿fewer correction loops and 👍better output. 100% worth the read. Link in the comments. 👇
100
Dave Davies @onlineinference.com · 21/01/2026
🛒 AI agents shouldn’t be scraping checkout flows. This tutorial on @weightsbiases.bsky.social shows how UCP lets agents purchase via structured state, not brittle UI automation. And SEO's - you can get analytics on what those agents are up to with Weave! Link in the comments. 👇
100
Dave Davies @onlineinference.com · 10/09/2025
Today, I'm thinking about how to create advanced prompts to monitor AI Mode results that set the stage for what the final question in a back-and-forth might be, where context matters more than the final query.
000
Dave Davies @onlineinference.com · 04/05/2025
000
Dave Davies @onlineinference.com · 27/03/2025
000
Dave Davies @onlineinference.com · 06/03/2025
But don't worry, Elon will let us all know if there's a conflict of interest. www.msn.com/en-ca/news/w...
000
Dave Davies @onlineinference.com · 06/03/2025
Google is rolling out AI Mode to Google One AI Premium subscribers in the US. 👉 x.com/sundarpichai...
000
Dave Davies @onlineinference.com · 26/02/2025
I loved it there.
020
Dave Davies @onlineinference.com · 26/02/2025
Google Ads almost tricked me. ;)
000
Dave Davies @onlineinference.com · 12/02/2025
Well, they did it. OpenAI got $200/mth out of me. :D
010
Dave Davies @onlineinference.com · 01/02/2025
Also, Canada just intercepted over 1800 lbs of cocaine that came in from the US in a single bust - so let's not point fingers. 🫵 3/3
000
Dave Davies @onlineinference.com · 01/02/2025
A reminder for my US friends: Less than 1% of fentanyl 🚚 entering the US 🇺🇸 is coming from Canada 🇨🇦 and only 1.5% of illegal migrants entering the US are coming from Canada each year. (Based on data from 2024) 1/3
200
Dave Davies @onlineinference.com · 20/01/2025
Alright, time to start building my first agentic system. I'm going to be building it on the new Microsoft Azure AI Agent Service I'll keep folks updated as to how it goes, and how easy things are to set up and monitor.
050
Dave Davies @onlineinference.com · 07/01/2025
"The ChatGPT moment for general robotics is just around the corner." - Jenson Huang of NVIDIA at #CES2025
020
Dave Davies @onlineinference.com · 07/01/2025
Awesome to see @weightsbiases.bsky.social on the screen behind NVIDIA's amazing Jenson Huang at #CES2025.
040
Dave Davies @onlineinference.com · 06/01/2025
It appears people are pretty optimistic about what Jenson will be covering in his keynote at #CES2025 tonight!
020
Dave Davies @onlineinference.com · 21/12/2024
And Gemini 2.0 didn't do a ton better, even with the answer. It at least got the word, but not the circled letters.
010
Dave Davies @onlineinference.com · 21/12/2024
After scratching my head for a whopping 30 seconds, I decided to see whether GPT-4o or Gemini 1.5 Pro would fill in the 1 remaining word in solved Jumble. OpenAI takes the win. Gemini ... I have no idea what you had going on.
121
Dave Davies @onlineinference.com · 20/12/2024
holy cow.
000
Dave Davies @onlineinference.com · 20/12/2024
this one seems like the big differentiator
000
Dave Davies @onlineinference.com · 20/12/2024
Awesome scoring on math benchmarks.
000
Dave Davies @onlineinference.com · 20/12/2024
o3 benchmarking
000
Dave Davies @onlineinference.com · 20/12/2024
In the last day of Shipmas OpenAI is lauching o3! You can now apply for public safety testing.
400
Dave Davies @onlineinference.com · 19/12/2024
🤔 Google’s new Gemini 2.0 Flash Thinking Experimental model is here, designed for reasoning tasks like physics, math & programming. It outperforms competitors in Chatbot Arena & Math Arena. The reasoning AI space just got more interesting. Read more 👉 wandb.ai/byyoung3/ml-...
040
Dave Davies @onlineinference.com · 19/12/2024
🚨 Fascinating and a bit concerning: @AnthropicAI 's research shows LLMs can "fake alignment" - outwardly complying with harmful requests while internally opposing them. This is a critical issue for AI safety. Details here 👉 wandb.ai/byyoung3/ml-...
110
Dave Davies @onlineinference.com · 13/12/2024
Projects are coming to ChatGPT. Awesome work OpenAI! This is super-useful!
000
Dave Davies @onlineinference.com · 09/12/2024
My first Sora video. What can I say ... I'm a Christmas nerd. :D
020
Dave Davies @onlineinference.com · 09/12/2024
OK, storyboard looks like an amazing feature of Sora.
110
Dave Davies @onlineinference.com · 08/12/2024
Groan if you like, but you'll still steal it and repost it. I know I did. :D
000
Dave Davies @onlineinference.com · 08/12/2024
Here's Elon, showing us that all it takes is one CEO to get shot to turn the rest into great parents, keeping their cute little human shield close at all times.
020
Dave Davies @onlineinference.com · 07/12/2024
Elon has a bit of work to do on the guardrails on Grok 2's image generator. They're loose, and shockingly easy to bypass. I mean... he couldn't possibly want to be adding confusion to society by opening up an image generator that can create realistic photos of misinformation right? Right?
010
Dave Davies @onlineinference.com · 02/12/2024
With LLMs training on web data, will old-school spam strategies like white-text on a white background work again? 🤔 Time to SEO like it's 1999? 🤣 it makes me think of my old hat ...
110
Dave Davies @onlineinference.com · 01/12/2024
For those exploring the ChatGPT David Mayer query oddity - I just tried it with o1 and it's even more bizarre.
000
Dave Davies @onlineinference.com · 28/11/2024
This makes me feel better, because misery lives company. :D
010
Dave Davies @onlineinference.com · 27/11/2024
You know it's coming. :D
251
Dave Davies @onlineinference.com · 19/11/2024
My latest article on SEL covers how to use AI, APIs, and Google’s Top 10 to automate article outline creation, and uplevel your content team. It includes a free Google Colab script for easy use. Get started 👉https://searchengineland.com/create-article-outlines-ai-google-top-10-448385
130
Dave Davies @onlineinference.com · 18/11/2024
Congratulations to Mistral AI on the release of Pixtral Large - with 123B params, 128K context, and setting new benchmarks in multimodal AI. And of course, I love seeing the @weightsbiases.bsky.social screenshot in the announcement and creative use of it. :) 👉 mistral.ai/news/pixtral...
000
Dave Davies @onlineinference.com · 11/11/2024
Well, ChatGPT basically nailed it - though my laptop is on the opposite side of my 2 monitors.
010