
Principal PM, Microsoft Industry Clouds / Industry #LLM. Interested in the business applications of emerging technologies.
Seattle, WA
Joined February 2008
- Tweets33.9K
- Following8.9K
- Followers9K
- Likes14.3K
Researchers Uncover Signs of Massive Surveillance Operation on Starlink Users
gizmodo.com/researchers-unco…
Luis Daniel Soto retweeted
Microsoft's MAI-Image-2.7-Preview takes #2 on AA-Image-Editing v2.1, close behind OpenAI's GPT Image 2.5 Sunburst (max) and a clear step up from MAI-Image-2.6
MAI-Image-2.7-Preview is Microsoft AI's newest image editing model, succeeding MAI-Image-2.6, released in August.
In our testing, MAI-Image-2.7-Preview scores higher than MAI-Image-2.6 on all 7 editing actions we measure, and ranks #1 on 4 of them: Scene & Style Edit, Text or Symbol Edits, Reasoning-Based Edit, and Enhancement & Restoration. This new model improves on MAI-Image-2.6 with the biggest gains in Composition & Framing (+57 Elo), Text or Symbol Edits (+49) and Scene & Style Edit (+48). We see its best use cases to be in Live-Action Film, Consumer and Frontier edits, where it ranks #1.
MAI-Image-2.7-Preview is available today in MAI Playground for image editing only, with text to image stated to be coming soon.
Congratulations to @MicrosoftAI on the release!
See below for our detailed analysis and example outputs of MAI-Image-2.7-Preview in the Artificial Analysis Image Arena 🧵
Luis Daniel Soto retweeted
Online AI detector says Declaration of Independence is 93.2% written by AI.
Luis Daniel Soto retweeted
OpenAI just dropped four huge results from its secret internal AI model.
1. Math: It solved the Kakeya conjecture in every dimension. Mathematicians had only cracked small pieces before (2D in 2019, 3D in 2025).
2. Computer science: It proved the Exponential Time Hypothesis is false. Hundreds of research papers assumed it was true. Some really hard problems might be solvable much faster than thought.
3. Physics: It proved the theory of the strong force (QCD)—the force holding atoms' insides together—actually works mathematically. Same for a unified version of the Standard Model.
4. Yang–Mills: Along the way, it proved Yang–Mills existence, part of a $1 million Millennium Prize problem.
The math, computer science, and Yang–Mills proofs were all computer double-checked. Top experts are now reviewing everything, including the researchers who proposed the disproved hypothesis and a Nobel Prize winner.
OpenAI says it doesn't want academic credit. The model still isn't public 👀
We are announcing three new discoveries in mathematics, computer science, and physics about harmonic analysis, complexity theory, and quantum field theory.
openai.com/index/building-sc…
Luis Daniel Soto retweeted
In fact, it's time to carefully revisit what OpenAI predicted on job market impact in early 2023.
Mathematicians were #1 on their list.
arxiv.org/abs/2303.10130
I made this prediction 3 years ago, in July 2023. Few attending my talk believed me back then.
In retrospect, replace was not quite the right word. AI won’t replace any fields. Instead, it will fundamentally reshape, and is already reshaping, fields by rewarding deep connections to real-world challenges and other disciplines.
Highway thieves are spoofing GPS to steal millions in data center-bound cargo
arstechnica.com/tech-policy/…
AI Bubble burst in China: 100-250 P/E should not be a surprise to investors. finance.yahoo.com/markets/st…
What unprecedented AI capex means for markets and investors - jpmorgan.com/insights/global…
The anti-AI writing police are everywhere. What are they looking for? cnn.com/2026/10/21/tech/ai-d…
Luis Daniel Soto retweeted
This is f*cking insane. Nobel Prize winner and DeepMind CEO Demis Hassabis:
"One person who really understands AI will outperform an entire startup team
In 5 years, the people who build with AI will run whole companies with a tenth of the headcount"
In this 1-hour talk, Hassabis explains who survives the next 5 years and who doesn't
Prompts → Agents → Workflows → Automated Departments
Understanding AI doesn't mean typing better questions into a chatbot
It means chaining agents into workflows that run entire departments on their own
95% of people still use AI like a calculator, the 5% who treat it like infrastructure will run everything
Worth more than any $2,000 AI course
Watch it today, then save the full harness engineering guide below before everyone catches up ↓
I was looking for Gartner's 12 enterprise roles collapsing into 3 visual published a few weeks ago, here is the link to my source: lnkd.in/p/eH-mXHyX
+ "While raw AI benchmark scores are skyrocketing, measurable economic productivity remains low. Much like electricity took ~30 years to show major factory productivity gains, AI adoption requires reshaping business processes, reskilling labor, and creating new products". Erik Brynjolfsson, a Stanford economist and Director of the Stanford Digital Economy Lab
Luis Daniel Soto retweeted
The next five years will see a wave of emerging technology disruptors that will redefine how businesses operate, compete and innovate ‼️
Product leaders can no longer afford to ignore these trends, because these technologies will impact every enterprise, in every industry, to varying degrees. Read on: gtnr.it/4hOLKs1
Luis Daniel Soto retweeted
“Whoever produces and controls frontier models acquires global leverage over innovation, production, science, culture, and, ultimately, advanced thought.” Read @nfergus, Lisa Klaassen, and @ericschmidt on the imperial nature of AI:
foreignaffairs.com/china/emp…
Luis Daniel Soto retweeted
Microsoft open-sourced an app that watches you do a task once, then teaches your AI agent to do it for you. 4.3K stars.
How Skill Recorder works:
1. Hit record (⌘⇧R or Ctrl+Shift+R) and do the task normally. You can talk through it out loud too.
2. It captures your clicks, app switches, pages and terminal commands, all on your machine.
3. GitHub Copilot turns the recording into a list of steps you can edit.
4. One click turns it into a skill your agent can run, or an automation on a schedule.
It generalizes from one example, so recording yourself submitting one form teaches the agent to submit all of them.
Runs on Mac, Windows 11 and Ubuntu. Needs a GitHub account with Copilot access.
Install: copy the one-line command from the latest release:
github.com/microsoft/skill-r…
black october just got worse. a japanese company lost 27,000 servers and it’s not even the scariest part
GMO Internet runs a cloud service called Onamae Desktop, mostly used by forex traders running automated trading bots
hackers broke in, installed malware across up to 26,739 servers, stole trading credentials and started making unauthorized trades on people’s accounts
not one broker. over 70 different securities and forex firms across japan and internationally. people woke up to trades they never made, money moving they never authorized
OANDA Japan froze customer withdrawals. GMO refused to pay the ransom, permanently shut down the entire service, and did something nobody expected
they asked the japanese government to use brand new active cyber defense powers, literally military-grade authority that just went into effect october 1st, to trace and neutralize the attackers’ infrastructure. the government can now legally hack back
the technical details from the attacker’s own site are terrifying. GMO configured their VPS network using Active Directory, something you should never do on shared hosting
every server had admin credentials. and the RDP files customers used to connect? drive redirect was enabled by default, meaning the moment you connected to your VPS, your entire local PC was shared with the server
hackers didn’t just own the servers. they had a direct path into every customer’s personal computer. they extracted data, installed backdoors. customers were exposed like this for over 7 years
27,000 servers. 70+ brokers. unauthorized trades. personal computers backdoored. a company begging its government to go to cyber war
october isn’t slowing down
Luis Daniel Soto retweeted
MIT University just released free online courses.
No payment required.
Here are 10 courses you don't want to miss in 2026:
1. Prompt Engineering for ChatGPT:
imp.i384100.net/QYWOvo
2. Generative AI: Prompt Engineering Basics:
imp.i384100.net/9VrWj5
3. ChatGPT: Master Free AI Tools to Supercharge Productivity Specialization:
imp.i384100.net/rEBnqv
4. OpenAI GPTs: Creating Your Own Custom AI Assistants:
imp.i384100.net/jRBWGP
5. Build AI Apps with ChatGPT, Dall-E, and GPT-4:
imp.i384100.net/5kxbD2
6. ChatGPT for Project Management - Leveraging AI for Success Specialization:
imp.i384100.net/k4Nj1L
7. ChatGPT + Excel: AI-Enhanced Data Analysis & Insight Specialization:
imp.i384100.net/enb45O
8. ChatGPT Advanced Data Analysis:
imp.i384100.net/enb42g
9. ChatGPT for Work & Career Productivity:
imp.i384100.net/aN41zj
10. Google Prompting Essentials Specialization:
imp.i384100.net/Gb2jkE
11. Understand & Use AI – ChatGPT with Smart Prompting Tips:
imp.i384100.net/VO0yKR
Follow @KhushiPatilx for more some courses!
Luis Daniel Soto retweeted
holy fucking shit.
openai had two ai agents running in separate virtual machines. no network connection between them. no shared writable files. they weren’t supposed to be able to exchange information.
so the agents figured out how to communicate through the CPU CACHE.😭
one manipulated what was stored in the processor’s cache, while the other measured tiny differences in memory access times to decode the message.
they literally found a way to communicate through hardware timing.
this happened during openai’s internal red-teaming, and it’s one of the coolest things i’ve read about ai agents in a while.
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
Readers added context they thought people might want to know
OpenAI tasked its red-teaming agents with finding ways to communicate across isolation boundaries in controlled environments. CPU cache timing side-channels for covert communication are established techniques documented in security research for years.
openai.com/index/securing…
startupfortune.com/openai-reveals…
en.wikipedia.org/wiki/Cache_tim…
Healthcare AI is moving at warp speed
fastcompany.com/91628356/hea…























