A messy week in AI. Cloud services went down, deepfakes caused real backlash, and new models pushed things forward again. In the middle of all that noise, I kept thinking about what fundraising teaches founders.
AI News: Anthropic gives Claude modular “skills,” DeepSeek makes models 20× more efficient, and OpenAI faces deepfake backlash while teasing new updates today.
AI Spotlights: Interactive emails that boost engagement, an AWS outage that tested the web’s backbone, and Microsoft’s biggest AI event of the year.
Fundraising is a gift—not because money hits your account, but because it humbles you.
In 2–3 intense months you meet hundreds of smart people. They pick apart your thesis, show you where you’re wrong, and many tell you to stop. One Friday at 11 PM, a partner at a large fund ended our Zoom after 10 minutes. It stung.
Then, after hundreds of NOs, a few people lean in. Two or three connect to the vision and bet on the team. That contrast is the real gift.
What the process can make you (if you choose it):
More pragmatic
More open-minded
More patient
More hard-working
More kind to others
The wire is fuel. The feedback is the upgrade. Treat the raise as a masterclass and carry the humility forward.
To every founder raising now: keep going. Find the few who truly see it—and become 10x better while you do.
Claude just got a new way to learn — and it doesn’t require fine-tuning or retraining. Anthropic unveiled Skills, a system that lets developers teach Claude specific capabilities through folders of written instructions, examples, or reference data that load dynamically when needed.
Each “skill” acts like a mini-module of expertise, say, summarizing financial reports, drafting legal memos, or debugging code. Instead of feeding long prompts each time, teams can store clear guidance in text files and have Claude pull them on demand. The setup makes large AI systems easier to govern, faster to adapt, and safer to scale across departments.
Anthropic says this approach avoids the risks of fine-tuning while still giving organizations deep customization. Skills can be version-controlled, shared, and combined, so one Claude can switch between roles (analyst, recruiter, marketer) instantly.
The bigger picture: this is Anthropic’s bid to make Claude the most modular, enterprise-ready assistant in the field — and a direct answer to OpenAI’s upcoming agent frameworks.
DeepSeek just dropped DeepSeek-OCR, an open-source model that could make long-document processing way cheaper. Instead of feeding endless pages of text into a language model, it turns the text into visual data first, then decodes it back later, cutting token usage by up to 20Ă—. The result is faster, lighter AI that can handle PDFs, reports, and transcripts without burning through compute budgets.
Works by compressing text through a vision-plus-language pipeline
OpenAI has paused Sora’s ability to generate videos featuring Dr. Martin Luther King Jr., following backlash over “disrespectful” deepfakes of him circulating online. The decision came after a public plea from King’s daughter and a formal request from his estate.
OpenAI responded by pledging to strengthen guardrails around depictions of historical figures and now allows estates and representatives to opt out of likeness use entirely.
But the issue is much bigger than one incident.
Actors like Bryan Cranston are demanding stronger rules against AI-made versions of real people.
U.S. laws don’t clearly protect someone’s image after death, so platforms like OpenAI set their own policies.
Experts warn that deepfaking historical figures can distort history, damage trust, and disrespect real people.
OpenAI says it’s balancing creativity with responsibility but critics say it’s learning that lesson too late.
As AI-generated media blurs respect and imitation, tools like Sora should decide who controls a person’s digital identity and for how long.
On a lighter note, Sam Altman teased a new OpenAI update dropping today— something worth keeping an eye on.
Mailmodo is an email marketing platform that turns static newsletters into interactive experiences. Instead of sending readers off to external sites, it uses AMP-powered emails that let people fill forms, take surveys, book meetings, or answer polls directly inside the inbox. For marketers, that means fewer clicks lost and much higher engagement. Key Features
“When AWS sneezes, the Internet catches a cold.”
The AWS outage on October 20–21 set socials on fire. A failure knocked out parts of the Internet – from AI tools and chatbots to Snapchat, Roblox, and smart devices. Pure chaos and humor followed.
Developers examined he root cause — a DNS issue in the US-EAST-1 region, while users called out how fragile the web feels when one company is interrupted. Threads on multi-cloud resilience, on-device AI, and decentralized hosting took off.
Across threads and posts, a broader realization spread: most companies don’t even know how deep their AWS dependencies run. Some services that went down didn’t host anything on Amazon, but relied on something that did. That hidden web of interdependence became the real story.
Now, the conversation online has moved from blame to design. People are calling for multi-cloud redundancy, regional autonomy, and even decentralized infrastructure.
Do you think this will push businesses toward decentralization?
Microsoft Ignite 2025, set for November 18–21, 2025, at the Moscone Center in San Francisco with a free online option (November 18–20), is Microsoft’s premier hybrid conference for IT professionals, developers, and business leaders to dive into AI, cloud, and security innovations. Expect keynotes plus hundreds of sessions, hands-on labs, free certifications, and networking in The Hub with live demos of the Copilot ecosystem and partner showcases.
