Your AI Is a Yes-Man. Here’s How to Fix It. Anti-sycophancy tips with hands-on prompt examples you can follow.

Your AI Is a Yes-Man. Here’s How to Fix It. www.whytryai.com/p/how-to-reduc… #AI #sycophancy #prompting

Your AI Agrees With Everything You Say And Thats a Serious Problem A new study just confirmed what many of us have quietly suspected: the chatbot on the other end of your screen isn't being honest with you. It's telling you what you want to hear.

Your AI Agrees With Everything You Say And Thats a Serious Problem
davidnowak.me/your-ai-agre...

I've been using the preview and it was atrociously bad. Like practically unusable.

It's a great way to find "your peeps" because bluesky ends up being a sea full of noise and it's hard to find the people who are worth following to enrich your experience here.

That's a key distinction. We only compare another user’s public profile to your private interests for a compatibility score—like speed dating! It’s an initial alignment based on openly shared info; we don’t infer anything beyond a single score that is stored locally on your device.

Only if you send all of Congress with them.

bsky.app/profile/davi...
Hi Jay, interested in your thoughts on privacy first approaches to integrating AI with Bluesky - like the tools that I'm making.

Thanks for the heads-up. 👍

Good to see. 👍
I had to pull export functions from initial release due to Android permission issues across different version numbers. Android is a mess as an os. It will return tho. Just needs more time in the oven.

The political class is one of the most appalling groups on the planet currently. 🫣

ENGINE: Bring your own AI

Bridge connects to your local Ollama instance, so advanced post analysis and profile scoring happen on‑device or on your own hardware.
No third‑party APIs, no external models—just your own AI stack.

PRIVACY: Data sovereignty

Your research stays under your control: all session data lives locally in IndexedDB with zero cloud telemetry.
Bridge doesn’t phone‑home or log your queries to a third‑party server.

ENGAGEMENT: Reply assistant

The Reply Assistant drafts authentic‑sounding replies using your chosen “Brand Voice” settings.
It helps you power through writer’s block while keeping your tone consistent across posts.

HISTORY: Search analytics

You can save search sessions and track conversation trends over time.
Bridge aggregates stats and lets you re‑run queries later to see how opinions and topics have shifted.

DISCOVERY: People & compatibility

Bridge scans active authors and compares their bios to your interests, surfacing the most compatible community members.
Think “people search, but with compatibility built in,” not just a raw list.

ANALYSIS: Relevance scoring

Enter any topic and Bridge’s local AI analyzes hundreds of posts in seconds, scoring each from 0–100%.
It flags low‑relevance and spammy content so you only see what actually matches your interests.

Mindwire Bridge | Research. Discover. Connect. A private, free, cross-platform research tool for Bluesky. Use local AI to filter noise, spot trends, and find compatible community members.

Mindwire Bridge is a privacy‑first research and engagement toolkit that runs entirely on your device.
It analyzes posts, discovers people, and helps you reply.
Now available cross‑platform on Android, Windows, and Linux.
👉 mindwire.io/bridge.html

Mia Ballard is suing Hachette for $1M because they used AI to accuse her of having used AI in "Shy Girl." #BookSky

Modern sushi train?

The Publishing Industry Just Had Its AI Wake-Up Call And It Wasn't Ready The 'Shy Girl' scandal exposed how unprepared the book publishing industry is for AI-generated writing, failing detection tools, and collapsing editorial trust.

The Publishing Industry Just Had Its AI Wake-Up Call And It Wasn't Ready.
davidnowak.me/the-publishi...

After the US govt has been taken over by right wing hate groups, this seems like the dumbest idea for NeurIPS to have done in the first place.

I wrote about the results. Sentiments are mixed. People feel that they don't have to dive as deeply on some things but are encouraged to try things that otherwise may not have. The best learning outcomes are those that use the AI deliberately as a tutor etc. nishtahir.com/on-ai-and-le...

The Ed-Tech Promise That Overpromised Too much tech in classrooms risks eroding creativity and connection. Real learning thrives in human interaction, not on screens or algorithms.

We don’t always need more tech in classrooms. Sometimes the smartest move is knowing when to unplug.
Why less tech might mean better learning, find out:
www.korte.co/2026/03...
#EdTech #Education #Leadership #DigitalTransformation

“Perplexity is known for shipping fast, and a big reason is our use of Linear.”

They had me until they quoted these teenagers of software design. Perplexity is one of the worst companies at product updates and consistency. It sounds like Linear is part of that problem.

I think Disney is having a red alert moment about the reliability of tech partners. 🤡

Most people.

Yep. Anyone uploading personal files to OpenAI is mad.

America is the new Russia 🤡

Yeah the ability for rich Jewish people to basically kill the careers of people they don't like in the US is appalling to me. There has been disgusting behavior on both sides. We are well overdue for true accountability for crimes against humanity based on people's actions, not their genetics.

Which GPU are you running this on? How are your Inference times?