This is an actual line that was added to the official system prompt for Codex for GPT-5.5 by OpenAI. Usually the system prompt is as minimal as possible, so I assume it would otherwise mention goblins a lot.
AIs are weird.
AIs are weird.
Generate viral LinkedIn posts in your style for free.
Generate viral LinkedIn posts in your style for free.
The best LinkedIn Posts
Explore the top viral LinkedIn post examples, trends and ideas from the best LinkedIn influencers.
LinkedIn Posts that went viral yesterday
How to get professional headshots with ChatGPT.
1. Upload 4 simple photos of you.
2. Make sure to turn on “Thinking” in ChatGPT 5.5.
3. Copy & paste the following prompt:
"a corporate headshot of this person"
Or to be more specific, copy paste this prompt:
"Use the attached image as the exact facial reference. Create a hyperrealistic professional studio editorial portrait that preserves the person’s facial identity, proportions, expression, and natural features.
Portrait style: contemporary high-end photography.
Framing: medium head-and-shoulders shot, 4:5 aspect ratio.
Pose: direct gaze into camera, neutral expression.
Background: pure uniform black, no texture, no visible environment.
Lighting: soft studio lighting with a side key light and gentle fill. Add subtle side chiaroscuro for depth and separation from the black background.
Skin: natural, realistic skin texture with fine detail. No smoothing, no plastic effect, no artificial shine.
Eyes: very sharp focus, crisp catchlights, high detail.
Lens look: 85mm portrait lens, shallow depth of field.
Color: neutral color correction, natural tones.
Output: high-resolution, clean editorial portrait.
Negative prompt:
No caricature.
No painting or illustration style.
No fake skin.
No facial distortion.
No exaggerated smile.
No glossy or artificial skin.
No visible background elements.
No texture in the background.
No overprocessed retouching."
ChatGPT released a new '5.5' model, which is better at images. Actually, the best at images.
Then search, and coding, and writing.
This is a new kind of model. Definitely a leap.
I just wrote ChatGPT image guide with copy-and-paste prompts like this: https://lnkd.in/dgHtVdBF.
♻️ Repost this to help others get cheap headshots.
1. Upload 4 simple photos of you.
2. Make sure to turn on “Thinking” in ChatGPT 5.5.
3. Copy & paste the following prompt:
"a corporate headshot of this person"
Or to be more specific, copy paste this prompt:
"Use the attached image as the exact facial reference. Create a hyperrealistic professional studio editorial portrait that preserves the person’s facial identity, proportions, expression, and natural features.
Portrait style: contemporary high-end photography.
Framing: medium head-and-shoulders shot, 4:5 aspect ratio.
Pose: direct gaze into camera, neutral expression.
Background: pure uniform black, no texture, no visible environment.
Lighting: soft studio lighting with a side key light and gentle fill. Add subtle side chiaroscuro for depth and separation from the black background.
Skin: natural, realistic skin texture with fine detail. No smoothing, no plastic effect, no artificial shine.
Eyes: very sharp focus, crisp catchlights, high detail.
Lens look: 85mm portrait lens, shallow depth of field.
Color: neutral color correction, natural tones.
Output: high-resolution, clean editorial portrait.
Negative prompt:
No caricature.
No painting or illustration style.
No fake skin.
No facial distortion.
No exaggerated smile.
No glossy or artificial skin.
No visible background elements.
No texture in the background.
No overprocessed retouching."
ChatGPT released a new '5.5' model, which is better at images. Actually, the best at images.
Then search, and coding, and writing.
This is a new kind of model. Definitely a leap.
I just wrote ChatGPT image guide with copy-and-paste prompts like this: https://lnkd.in/dgHtVdBF.
♻️ Repost this to help others get cheap headshots.
I have been playing with the new Outlook agent that Microsoft announced today, and it is fine, but really awkward to use, since you have to ask for things in a chatbot window, then go to your drafts, etc.
The real problem is that it feels far behind already. Claude Cowork does the all the same functions (works with Gmail, too) and has better visibility across your life so its advice is sharper and the things it can do for you are much more advanced.
Microsoft owns the entire Outlook interface! Another chatbot on the side that can only take limited actions to help you feels like such a wasted opportunity, especially when both ChatGPT and Claude can already do a lot more.
The real problem is that it feels far behind already. Claude Cowork does the all the same functions (works with Gmail, too) and has better visibility across your life so its advice is sharper and the things it can do for you are much more advanced.
Microsoft owns the entire Outlook interface! Another chatbot on the side that can only take limited actions to help you feels like such a wasted opportunity, especially when both ChatGPT and Claude can already do a lot more.
Here is an AI trained just using text from 1931 or earlier, which leads to a lot of interesting experiments: can the model independently develop later inventions? Can it learn to code from examples alone? How does this
You can talk to the model here, which is a little like talking to someone from the past: https://talkie-lm.com/chat
Details of how it works and analyses here: https://lnkd.in/eeXRTFTh
You can talk to the model here, which is a little like talking to someone from the past: https://talkie-lm.com/chat
Details of how it works and analyses here: https://lnkd.in/eeXRTFTh
Your team doesn't need a babysitter. They need a leader.
If you don't trust your team to manage their own time, you either hired the wrong people or you're the wrong leader.
Most likely both.
You can't tell if someone is doing good work, so you measure what you can see: butts in seats, hours logged, bathroom breaks.
Result?
Your A-players quit first because they have options and working for you is exhausting.
Your B-players become C-players because you've trained them to wait for instructions.
Your C-players love it because now they can blame you when nothing gets done.
Here's what trust-based leadership looks like:
"I need this project finished by Friday. How you do it is up to you."
"Your quarterly goals are clear. Show me what you accomplished."
"You have a doctor's appointment? Handle your work and go."
Clear expectations. Total autonomy. Real accountability.
No timesheets. No status meetings about status meetings. No asking permission to live your life.
Just adults doing the work they were hired to do.
You hired your people for a reason. Let them reason.
-DM
If you don't trust your team to manage their own time, you either hired the wrong people or you're the wrong leader.
Most likely both.
You can't tell if someone is doing good work, so you measure what you can see: butts in seats, hours logged, bathroom breaks.
Result?
Your A-players quit first because they have options and working for you is exhausting.
Your B-players become C-players because you've trained them to wait for instructions.
Your C-players love it because now they can blame you when nothing gets done.
Here's what trust-based leadership looks like:
"I need this project finished by Friday. How you do it is up to you."
"Your quarterly goals are clear. Show me what you accomplished."
"You have a doctor's appointment? Handle your work and go."
Clear expectations. Total autonomy. Real accountability.
No timesheets. No status meetings about status meetings. No asking permission to live your life.
Just adults doing the work they were hired to do.
You hired your people for a reason. Let them reason.
-DM
It’s time we understand this truth! Spending all our days on the 0000000000000.1% and not acknowledging how wonderful most people are ❤️❤️❤️❤️❤️❤️🔑
Share this with a cynic or someone who needs a reminder 💖🔐
Share this with a cynic or someone who needs a reminder 💖🔐
iPhone killer? OpenAI is reportedly developing an AI-first smartphone to compete with Apple’s mobile dominance 😳
According to a new supply-chain report from Ming-Chi Kuo, OpenAI is working on an AI-first smartphone with MediaTek, Qualcomm, and Luxshare.
Mass production is reportedly targeted for 2028, with specs and suppliers expected by late 2026 or Q1 2027.
The idea is simple: OpenAI is trying to turn the smartphone from an app launcher into an agent launcher 🤖
Today, your phone is built around icons:
Banking app.
Calendar app.
Uber.
Gmail.
Slack.
Maps.
OpenAI bets that the next interface is not “open the app” → It’s “do the thing.”
↳ Custom SoCs optimized for AI agents
↳ On-device small-model execution + cloud offload
↳ Continuous context from sensors, location and usage
↳ OpenAI subscriptions bundled into hardware
↳ A developer ecosystem built around agents, not apps
This is why the phone matters.
Agents are weak when they live inside someone else’s sandbox.
iOS and Android were built for apps, permissions, notifications, and controlled access.
An AI agent wants memory, context, tools, sensors, identity, payments, and permission to act.
That is an operating system problem.
Not a chatbot problem.
My takeaway here is this:
It’s not about whether OpenAI can build a phone. They definitely can.
It’s about whether people will switch.
Because ecosystems are sticky.
Carrier deals are tough.
Hardware is hard for a reason.
And most importantly, Apple and Google will not sit still.
P.S. check out How to Build an AI Agent from Scratch (With Working Code) 🤖: https://lnkd.in/dP6r5gHb
According to a new supply-chain report from Ming-Chi Kuo, OpenAI is working on an AI-first smartphone with MediaTek, Qualcomm, and Luxshare.
Mass production is reportedly targeted for 2028, with specs and suppliers expected by late 2026 or Q1 2027.
The idea is simple: OpenAI is trying to turn the smartphone from an app launcher into an agent launcher 🤖
Today, your phone is built around icons:
Banking app.
Calendar app.
Uber.
Gmail.
Slack.
Maps.
OpenAI bets that the next interface is not “open the app” → It’s “do the thing.”
↳ Custom SoCs optimized for AI agents
↳ On-device small-model execution + cloud offload
↳ Continuous context from sensors, location and usage
↳ OpenAI subscriptions bundled into hardware
↳ A developer ecosystem built around agents, not apps
This is why the phone matters.
Agents are weak when they live inside someone else’s sandbox.
iOS and Android were built for apps, permissions, notifications, and controlled access.
An AI agent wants memory, context, tools, sensors, identity, payments, and permission to act.
That is an operating system problem.
Not a chatbot problem.
My takeaway here is this:
It’s not about whether OpenAI can build a phone. They definitely can.
It’s about whether people will switch.
Because ecosystems are sticky.
Carrier deals are tough.
Hardware is hard for a reason.
And most importantly, Apple and Google will not sit still.
P.S. check out How to Build an AI Agent from Scratch (With Working Code) 🤖: https://lnkd.in/dP6r5gHb
What doesn’t work on LinkedIn is spamming people through messages.
The thing that blows me away is, so many salespeople and entrepreneurs get spammy LinkedIn messages and hate receiving them — yet they’re so willing to spam other people.
There’s a lot of tone deafness and hypocrisy there because the truth is, it’s just not the best way to build a following and extract value out of LinkedIn.
The proper way to do it is by giving value to the community first by producing or engaging with content.
The thing that blows me away is, so many salespeople and entrepreneurs get spammy LinkedIn messages and hate receiving them — yet they’re so willing to spam other people.
There’s a lot of tone deafness and hypocrisy there because the truth is, it’s just not the best way to build a following and extract value out of LinkedIn.
The proper way to do it is by giving value to the community first by producing or engaging with content.