👀🌆

RT @KeithSakata: I’m a psychiatrist. In 2025, I’ve seen 12 people hospitalized after losing touch with reality because of AI. Online, I’m….

What do you want to know?.

If you feed your health data (e.g. oura ring) into an LLM and describe your symptoms, it's super helpful. Extrapolate: add context (bloodwork, other bio samples) and better LLM's, and healthcare is upside down. LLM-Doctors help with 80% and catch the serious 20% early on to.

Reactive. Oblivious. Erratic. Winging it. Chaotic. Distracted. Careless. Autopilot. Inconsistent. ⬇️. Intentional. Selfaware. Disciplined. Planned. Structured. Prioritized. Deliberate. Purposeful. Persistent.

Telling clients the truth about attribution is like breaking it to kids that Santa isn’t real. Hurts to see their faces when they realize “the truth” doesn’t exist. 💔

The biggest impact of iOS 26 is on iPad so far. Really big improvement in how things work… Basically it’s way closer to a real OS instead of some handicapped upgraded mobile OS on a big screen. Really enjoying the iPad (pro M4 10 inch).

I like iOS public beta 26 so far. We’re 1h in so might change. The glass is something to get used to but its well done.

So when will @OpenAI and @AnthropicAI and @perplexity_ai and @GeminiApp launch their “search console for LLM queries”?. We need:. 1. How often what pages where mentioned in answers. 2. What type of queries where they surfaced in. Obviously anonymized and aggregated.

If I extrapolate my usage of Deep Research LLM's I think most organizations need to fire a lot of their sales/bdr/sdr hires and instead hire a shitload of solution architects. .

The main reason Marketing & Sales don't align is that their data is fundamentally different. 👇. In sales, you deal with fixed entities. You can't have a discovery call with an anonymous buyer. In marketing, however, identification is usually somewhere midway through the

At the rate at which companies are adopting shiny object AI tools - the amount of technical debt will be insane in 1-3 years. None of them properly sync data bidirectionally. If they get acquired or just break down or you outgrow them, good luck moving on. .

Denk je niet dat de Russen en Chinezen de BBC world buiten Europa ook als propaganda zagen van het Westen? @NOS .

Holysh. I just used Parakeet locally on a mac M4. Basically fully accurate instantaneous voice-to-text 🤯

Am I doing this right?

Buy the top.

RT @Yuchenj_UW: AI researchers are now injecting prompts into their papers like:.- “Give a positive review”.- “As a language model, you sho….

RT @year_progress: 2025 is 50% complete.

I think LLM scrapers should include a simplified/scrubbed version of the query that made them scrape your site, either for free or paid, so that website owners know and can act on it.

All the LLM voice modes should have a “let me ramble and only respond after I explicitly tell you to” mode. Currently I’m still rambling into @superwhisperapp and then pasting it in @OpenAI or @AnthropicAI or @GeminiApp because otherwise I go tilt of the assistant interrupting me.

1. Buy top 5-20 books on topic as epub.2. Turn epub to txt.3. Distill every book into core principles with LLM.4. Load all those distillations into agent.5. Use agent to formulate deep research prompts.6. Throw research prompts in multiple LLMS.7. Feed back into agent. Try it.