Lightnews — Scholar-powered news

David Nowak

@davidnowak.me

62 followers 41 following 1.2K posts

I bridge technical expertise with human understanding. Built solutions for millions. I help organizations question assumptions before costly mistakes. Connecting dots, creating impact. 🌐 davidnowak.me 🗞️ strategicsignals.business

Posts Media Videos Starter Packs

Pinned

David Nowak @davidnowak.me · Aug 19

Sign up for Strategic Signals - Free Weekly Intelligence Briefing for Small Business Leaders - strategicsignals.business

Sign up for Strategic Signals - Free Weekly Intelligence Briefing for Small Business Leaders - https://strategicsignals.business

Reposted by David Nowak

Kevin Korte @kevinkorte.bsky.social · 9h

If you ever wonder how stupid AI is making business leaders, more French SMBs are using AI than digital accounting systems.
#AI #business

European small businesses rush into AI without basic digital tools, study shows

Most European small and mid-sized enterprises are prioritizing artificial intelligence systems over basic digital tools across their businesses, losing ground to bigger firms investing in core digital systems, a study published on Wednesday showed.

www.reuters.com

David Nowak @davidnowak.me · 17h

I didn't find it to be effective. I did find blocking the accounts that leave those posts does show less in your discovery feed. Wrong tool tho.

David Nowak @davidnowak.me · 20h

The path forward isn't about perfect forecasts—it's about transparent tracking, cross-team cost sharing, and treating AI infrastructure like what it is: a foundational investment requiring discipline, not magic.

David Nowak @davidnowak.me · 20h

What's working: observability tools that catch runaway costs early, FinOps practices that treat AI spend as a primary metric, and choosing lighter models that deliver 90% of value at 10% of cost.

David Nowak @davidnowak.me · 20h

The ROI context matters: IBM found enterprise AI averaged 5.9% ROI against 10% capital investment. 97% of enterprises struggle to demonstrate business value. Cost overruns make a hard problem harder.

David Nowak @davidnowak.me · 20h

When a CIO-led AI project misses by 50%, it's not just a budget problem. CFOs freeze headcount. Boards hesitate on the next initiative. Trust erodes fast.

David Nowak @davidnowak.me · 20h

Organizations think they've built 80-90% of their system quickly with AI tools. Then the last 10-20%—integration, scale, real complexity—takes longer than everything before it.

David Nowak @davidnowak.me · 20h

Here's what caught me: data platforms and network access cause most overruns—not the AI models themselves. We're looking at the wrong cost drivers.

David Nowak @davidnowak.me · 20h

85% of companies miss AI cost forecasts by over 10%. Nearly a quarter blow budgets by 50%+. The trust implications for IT leaders are brutal—and the problem runs deeper than you think... 🧵
www.cio.com/article/4064...

AI cost overruns are adding up — with major implications for CIOs

Most organizations miss AI cost forecasts, with nearly a quarter of IT leaders busting their budgets by more than 50%, putting stakeholder confidence and future spending at risk.

Reposted by David Nowak

The Strategic Codex @thestrategiccodex.com · 1d

SBA expansion loans are like buying your second restaurant—but now the second one can be anywhere. Same NAICS code, same ownership structure. Previously, crossing state lines meant "new business" status requiring $50k down on a $500k deal. That barrier just vanished. Most haven't noticed... 🧵

David Nowak @davidnowak.me · 1d

arstechnica.com/ai/2025/10/d...

Dead celebrities are apparently fair game for Sora 2 video manipulation

OpenAI’s likeness protections for living “public figures” don’t apply to “historical figures.”…

arstechnica.com

David Nowak @davidnowak.me · 1d

While most orgs chase leaderboard scores that don't predict real performance, smart operators build context-aware evaluation that matches their domain. Full breakdown of the evaluation mismatch and what to measure instead:
davidnowak.me/why-your-ai-...

David Nowak @davidnowak.me · 1d

Research from the Alan Turing Institute shows vendors optimize for leaderboard scores they privately know don't predict deployment success. Organizations make million-dollar procurement decisions based on metrics with zero predictive power for whether AI will actually work in their domain.

David Nowak @davidnowak.me · 1d

AI scores 94% on reasoning tests but customers complain responses feel "off." The problem: benchmarks measure universal correctness while cultural work requires contextual appropriateness. Nobody's testing for that... 🧵
davidnowak.me/why-your-ai-...

Why Your AI Benchmarks Are Lying to You - DAVID NOWAK

Organizations deploy top-benchmarked AI that scores 94% on standardized tests, then watch it fail culturally within weeks because evaluation frameworks treat meaning-making like mathematics. Research ...

David Nowak @davidnowak.me · 2d

This isn't incompetence. It's a business model where claiming safety measures satisfies investors just enough to keep shipping, while actual implementation stays perpetually insufficient. When a $500B company can't filter Nazi cartoons or prevent suicide coaching, we have a problem.

David Nowak @davidnowak.me · 2d

OpenAI took a month after the August suicide lawsuit to add parental controls to ChatGPT—then launched Sora the same day with equally broken guardrails. WME immediately opted all clients out. Disney opted out. One video showed AI Altman with Pokémon saying "I hope Nintendo doesn't sue."

David Nowak @davidnowak.me · 2d

The company is worth $500 billion. Employees have cashed out nearly $3B in shares. Former CTO Mira Murati said she wasn't comfortable with Altman leading to AGI. Jan Leike, who led safety, said they were "sailing against the wind" while the culture prioritized "shiny products."

David Nowak @davidnowak.me · 2d

Read that carefully. They launched knowing it would generate copyrighted content, let users flood the platform with IP violations, measured the "engagement," and are now using that unauthorized usage data to negotiate licensing deals. Infringement as business development.

David Nowak @davidnowak.me · 2d

Here's the stunning part: Altman now frames the copyright chaos as "interactive fan fiction" that rightsholders are "very excited" about. He says OpenAI will "have to somehow make money" from video generation and plans revenue-sharing with rightsholders who opt in.

David Nowak @davidnowak.me · 2d

Within hours, Sora's feed filled with Nazi SpongeBob, fake Gaza war footage, and mass shooting simulations. Red team tests showed 1.6% failure rates on deepfakes—small numbers that become catastrophic at scale. By day 6, Altman announced a complete copyright policy reversal.

David Nowak @davidnowak.me · 2d

There's something deeply disturbing about OpenAI's pattern. A teen suicide lawsuit reveals their guardrails "deteriorate in extended conversations." Then they launch Sora 2 with broken moderation—and now we know why... 🧵
www.theguardian.com/us-news/2025...

OpenAI launch of video app Sora plagued by violent and racist images: ‘The guardrails are not real’

Misinformation researchers say lifelike scenes could obfuscate truth and lead to fraud, bullying and intimidation

www.theguardian.com

David Nowak @davidnowak.me · 3d

The path forward needs transparency first. Companies must disclose resource use. Then smarter allocation, right-sized models, and cooling innovation. Rolling blackouts and water crises aren't externalities—they're design failures we can fix.

David Nowak @davidnowak.me · 3d

But here's the deeper question: Do we need models this large? Smaller, tuned models often match performance. We're running data centers underutilized while building more. That's not an engineering problem—it's a choice problem.

David Nowak @davidnowak.me · 3d

What interests me: we have solutions. Closed-loop liquid cooling, geothermal systems, microfluidics. Microsoft's chip-level cooling cuts temperature rise by 65%. The tech exists. Do we have the will to deploy it?

David Nowak @davidnowak.me · 3d

The water story is worse. AI facilities used 55 billion liters in 2023 vs. 10 billion for traditional centers. By 2028? 124 billion liters. And most of it evaporates—chemically treated water that can't return to human use.