From Use, Not Theory
@AIFromUse
Former PM. | Trying to make sense of AI in public and share what works in real life in this new age
69% of pharma execs deploying marketing AI agents by end of year. challenge: making it work requires 'AI-ready data' most companies don't have yet
LLMs having 'aha moments' just like humans. didn't see that coming
People keep arguing about whether big language models can “really reason,” and a lot of that argument is rigged by a single pet test or narrow definition. A better way is to ask if they show the same functional signs of reasoning that we use to recognize reasoning in different…
stop waiting for the perfect tech stack your mvp doesn't need kubernetes. it doesn't need microservices. it doesn't need the latest js framework. it needs users who give a damn about your solution. build with what you know. ship fast. optimize later.
every AI marketplace should do this. tired of downloading "productivity tools" that mine crypto in the background
OpenClaw has partnered with VirusTotal, Google’s threat intelligence platform, to secure its skill marketplace ClawHub by integrating LLM-powered Code Insight scans. Every published skill is now scanned automatically for malicious behavior from known malware to suspicious code…
Nvidia supplier says AI isn't a bubble but isn't that exactly what someone with billions in AI orders would say?
This milestone highlights AI's potential in accelerating scientific discovery. However, we must consider how this impacts the role of human researchers in data analysis
Opus 4.6 is now better than human experts at reading and interpreting scientific charts
After 4 days of infrastructure chaos, Anthropic finally dropped it. Claude Opus 4.6 shows insane reasoning improvements + autonomous security flaw detection.
Great read! Impressive work from @LauraRuis
My PhD thesis is out 🥳🎓 How do LLMs, trained on trillions of tokens, reason? Can they generalise beyond their training data or are they constrained by what they've seen before? My takeaway: they can generalise beyond training in interesting ways, showing genuine reasoning
Claude sonnet 5 leaks looking wild 1M token context window. 80.9% on SWE-bench. half the price of opus. if these specs are real this changes everything for coding. anthropic might just eat OpenAI's lunch
realizing AI knows you better than you know yourself. your search history + chat logs = a mirror most people aren't ready to look into
In hindsight, the biggest AI shock won’t be technological. It will be psychological. What do you think people will struggle with most?
Apple adding Claude and GPT agents to Xcode isn't just convenience. It's the end of solo coding. When AI can explore your project, run tests, and fix bugs automatically, the barrier to building apps just disappeared. Expect an explosion of terrible apps by 2026.
Organizations need AI governance frameworks, data security protocols, and ethical guidelines before they need the latest models. Build the foundation first.
The vision of human-level machine intelligence laid out by Alan Turing in the 1950s is now a reality. Eyes unclouded by dread or hype will help us to prepare for what comes next go.nature.com/4qXTMRA
This Stanford paper marks the start of automated science. Today AI research, tomorrow drug discovery. When you can auto-generate hypotheses and test them in minutes, humans become the congestion. Machine-driven discovery is here.
New Stanford paper propose an automated executor that turns LLM research ideas into runnable code experiments and uses the results as feedback. It also warns that reward-based training can collapse into repeating small tweaks, so exploration needs active help. Instead of…
What's the best AI model today? Been testing different models for various tasks and curious what everyone's experience has been. Which one are you reaching for most often?
Insurance pouring money into AI but: 54% report poor AI outputs 35% admit data strategy comes first 24% have AI training programs Can't build reliable AI on bad data.
Gemini glasses this summer. Samsung + Warby Parker partnerships. Finally ditching the cyborg look for actual fashion. This changes everything
🚨 How is nobody talking about this?! Demis Hassabis just casually confirmed that Gemini-powered smart glasses are arriving "maybe by the summer." 🤯 He confirmed active partnerships with Samsung, Warby Parker, and Gentle Monster. This is huge because it means the "clunky"…
UK just announced Meta-backed AI team to upgrade public services Government AI transformation officially begins.
GPT-5.2 Pro: 31% on FrontierMath Tier 4. Previous record: 19%. Solved 4 problems that had never been solved by any model. We're watching AI get genuinely smarter, not just better at pattern matching.
New record on FrontierMath Tier 4! GPT-5.2 Pro scored 31%, a substantial jump over the previous high score of 19%. Read on for details, including comments from mathematicians.
"Elon Musk predicted AI smarter than any human by end-2026, surpassing collective human intelligence by 2030/31, at Davos." if ANY tech CEO thinks we're 2 years from AGI, we should be preparing for massive disruption
Apple vs OpenAI AI hardware race is heating up. Apple: AI pin with cameras (2027) OpenAI: AI earbuds (this year) One problem: Humane already proved nobody wants AI pins. techcrunch.com/2026/01/21/not…
United States Trends
- 1. Purdue N/A
- 2. #WWENXT N/A
- 3. Nebraska N/A
- 4. Jai Lucas N/A
- 5. Joe Hendry N/A
- 6. SWAT N/A
- 7. Real ID N/A
- 8. Cluff N/A
- 9. #LoveOthers N/A
- 10. Nancy Guthrie N/A
- 11. Hubert N/A
- 12. Baylor N/A
- 13. Jackson Drake N/A
- 14. Courtney N/A
- 15. RINO N/A
- 16. Matt Painter N/A
- 17. Alaska N/A
- 18. Rob Wright N/A
- 19. Fletcher Loyer N/A
- 20. #1000lbSisters N/A
Something went wrong.
Something went wrong.