Dare Obasanjo

AI companies: We’ve invented an unreliable guessing machine which sometimes doesn’t follow instructions. If you give it access to important systems it could kill us all. Also AI companies: Deploy this technology everywhere and replace your employees with it, otherwise China wins. IPO next month.

All of these parents should face charges for negligence. Every single one.

The economy is now so bad that MAGA can’t even afford to rent the libs.

Mathematicians and biologists used ChatGPT and Claude for their work only for those companies to later claim AI had discovered solutions to unsolved problems similar to their work. The concern that frontier labs will use your data to compete with you is sounding more like fact than fiction.

“Since you guys can’t figure out how to build a sandbox and are now threatening our bag, we’ll do it for you” - Nvidia paraphrased

By the time you realize you’ve made a strategic error, it’s often too late. America’s strategic error was electing a government that promised racism and incompetence because electing a Black woman was a bridge too far. It’s FAFO season, America.

Jerry Falade, a Black Nigerian writer, lost his $2m book deal because people mistook his writing for AI. As a fellow Nigerian, I totally understand. When I was at Georgia Tech, I remember a fellow teaching assistant who got upset with me because my writing style was so formal yet I spoke in slang.

In the future, historians will marvel at how a U.S. political party documented their project to destroy America (aka Project 2025), won the elections and set about to dismantle everything that made America great without any mainstream discourse or accountability. Fuck the Heritage Foundation.

VCs have pushed a long list of tech trends that were supposed to change everything but didn’t:

1. Daily deals (Groupon)
2. Consumer 3D printing
3. Micromobility
4. Coworking (WeWork)
5. Meal kits (Blue Apron)
6. NFTs
7. “Uber for X”

AI may be different. But hype isn’t evidence of inevitability.

"The second administration, I think overcorrected for the mistake of the first term with all these loyalty tests. You got an administration staffed by not very smart, but very loyal people." - Peter Thiel on the Trump administration and his cabinet of morons

So this is what it feels like to be embarrassed for an entire country.

Saturday Night Live skit of an interview with Anthropic CEO Dario Amodei.

Jane Smith who has been at the center of controversy for sharing an AI-generated photo wearing a “Lake America” sweatshirt with a group of others has lost her job as the CEO of Family First Credit Union. FAFO season continues for MAGA. https://yourhalifaxnow.ca/lake-america-sweatshirt-controversy-costs-michigan-credit-union-ceo-her-job/

Vercel’s AI Gateway data makes it hard to argue any AI lab has a durable moat. Anthropic fell from 84% → 37% of spend while OpenAI grew from 10% → 24%. Open-weight models grew from 1% → 29%, causing over half of Anthropic’s losses. Devs are increasingly willing to switch models when something better or cheaper arrives.

Going from claiming your company is working on technology that might destroy humanity to leaking you’re on track to generate $100 billion in sales this year, just in time for the IPO. Makes it hard to defend the claims that this hasn’t all been a cynical form of marketing and regulatory capture.

Apple continues to Think Different

I saw some people saying Meta’s Muse leaked someone’s address on Facebook Marketplace. The person worked with the Muse team to debug the issue. It turns out they did give Muse their address and permission to always act on their behalf but the gravity of the decision wasn’t clear enough to them

The risk with the AI boom is many things need to go perfectly or the bubble pops. While AI skeptics expect the bubble to pop because demand for AI will evaporate, I doubt that’s likely. It’s more likely that mundane logistics issues like opposition to data centers set of a domino effect of failures.

Claude Opus 5.5 was announced today and is the first model release since Anthropic called for pacing the frontier. I feel safer knowing Accenture consultants vetted that this model won’t destroy humanity.

Normally it would need to be a story about a police shooting to see the New York Times stoop to this level of euphemism and downplaying what happened.

Only 29% of Americans approve of the way Trump is handling his job as president according to a new poll by American Research Group, Inc.

As bad as things are now economically in the U.S., they are only going to get worse as long as Trump remains president. Between his ego and incompetence, there’s no clear path to rolling back tariffs or ending the conflict with Iran. Plan your life accordingly.

Anthropic’s latest threat report is wild. Claude was used to help build software for missile systems, autonomous drone swarms and cyberattacks against 50 organizations. But the craziest example may be one consultant using Claude to build a surveillance system covering 25 million SIM cards. AI isn’t just making businesses more productive, bad actors are also reaping the productivity gains of coding agents and AI as an orchestrator. https://www.anthropic.com/threat-intelligence-report-september-2026

“After careful consideration of Oracle's current business needs, we have made the decision to eliminate your role as part of a broader organizational change…As a result, today is your last working day.” - AKA our ̶e̶m̶p̶l̶o̶y̶e̶e̶s̶ data centers are our most valuable asset https://share.google/t5ymOHpAr8D55YTnV

I heard Converse posted a controversial ad and I wondered how bad it could be. Going in expecting something controversial I was still shocked by this…

"You got some MF who's made over $251B being a United States citizen that got here and stayed here illegally before he got his citizenship, sitting here lecturing us on who we should allow into the United States of America, and giving us a f**** Heil Hitler along the way" - Hunter Biden on Elon Musk

I’ve long noticed that people engage less if I post a link to a news story than a screenshot of it but assumed this was an algorithmic quirk of Threads. I hadn’t noticed the exact same thing happens on Bluesky as well but the data bears out. People simply engage with links less on social media.

Workers are suing Workday alleging its hiring software systematically discriminated against applicants. No rogue agent. No escaping a sandbox. Just software doing exactly what it was designed to do at scale. Another example of the sort of AI risk that needs to be focused on and legislated against. https://www.reuters.com/legal/government/these-workers-say-an-ai-hiring-platform-discriminated-against-them-who-should-2026-09-21/

One thing Obama said about AI that bears repeating is training agentic AI to hack websites and using AI to search for a cancer cure or solve the energy crisis are unrelated. Frontier lab CEOs treating these as intertwined are being disingenuous. It’s a choice to do both.

Imagine having such a racist crash out that this is how you had to be secured while the plane was diverted to drop your ass off with the cops. Don’t drink and fly if you’re this racist. https://www.wbtv.com/2026/09/05/passenger-arrested-after-being-restrained-with-duct-tape-zip-ties-during-american-airlines-flight-reports-say/

"If you are not confident about the safety of the products, and you're not confident in its functionality, capability, or safety, then just don’t release it.” - Jensen Huang, Nvidia CEO ~ Another common sense response to frontier labs abdicating responsibility for their rogue AI agents.

Every day of the war with Iran is Groundhog Day.

JP Morgan has implemented a $2,000/month token budget on some of its developers and is moving Claude usage to a sandbox versus running directly on developer machines. From tokenmaxxing to $24K/year is a steep drop but still implies ROI from AI is very high or budgets will need to drop further. https://www.businessinsider.com/jpmorgan-claude-spending-limit-security-engineers-2026-9

The most important skill in prompting is expertise in the domain you’re prompting for is well understood in many corners. It is why hiring across the software industry now favors super ICs and experienced seniors over entry level hires. AI makes you more productive if you know what you’re doing. https://www.seangoedecke.com/llms-reward-expertise/

Databricks has a great write up of best practices for reducing AI token costs.

1. Adopt open weight/lower cost models. Most tasks don’t need the best models.

2. Harnesses are a form of lock-in, adopt one which allows model flexibility.

3. Avoid hard per developer budgets. Prefer dashboards and spend gates.

https://www.databricks.com/blog/managing-ai-coding-costs-scale

Reading this excerpt about working at Lovable and the celebration of people becoming Super ICs instead of managers, what’s left unspoken is that you don’t need as many managers if you stop hiring junior people and have senior people do the grunt work. No managers because no one needs managing.

JD Vance’s midterm pitch to voters has extreme deadbeat ex-boyfriend energy.

AI driven layoffs are now a cliche and the memo from Expedia is no different. We get the mandatory “AI has radically changed what’s possible”, “we must fundamentally change how we work” and the new classic “organize into small end-to-end squads”. What’s novel is who is being let go; eight vice presidents. https://www.geekwire.com/2026/internal-memo-eight-execs-out-at-expedia-group-in-ai-driven-shakeup/

I saw someone post that Mamdani is solving problems they didn’t even realize could be solved by government. Contrast this with Trump and Republicans who only know how to create problems from starting unwinnable wars, rolling back food safety and destroying our healthcare system. We deserve better

You and I paying increased prices due to tariffs then corporations getting the refund is “trickle up economics.” Trump has invented even worse economic policies than Reagan. https://www.newsweek.com/tariff-refunds-split-companies-consumers-demand-share-12277265

Muse is now the top app in the App Store. My favorite use case was telling it to go find my unclaimed cash and do all of the related paperwork for me to claim it. I’m now up to $359 saved or earned by asking Muse to deal with tedious online tasks, I’ve been putting off. Disclaimer: I work at Meta.

Yelp and OpenTable have built-in AI chatbots but what I actually want is for them to expose an MCP endpoint so I can access them from my own agent. I understand why they don’t do it but I’d gladly pay a fee to make up for whatever they lose in app usage.

Anthropic’s models are today’s version of “Nobody gets fired for buying IBM”. It’s the safe choice, not the best one.

Daredevil while the events of Spider-Man: Brand New Day were unfolding.

Robinhood now makes more money from prediction markets than it does from stock trades or crypto. This adds weight to my theory that Bitcoin and crypto has crashed due to competition from Polymarket and Kalshi for the online gambler demographic.

Apple has accidentally become a major seller of AI servers, thanks to the popularity of the Mac Mini. OpenAI has reportedly bought tens of thousands, Anthropic rents them from AWS and Apple is now building software to cluster Macs together to run trillion-parameter open weight models like Kimi. https://www.theinformation.com/articles/apple-stumbled-ai-hardware-success-mac

Hank Green’s use of AI has been raised as an example of hypocrisy given Hank Green has been critical of the industry. However I think what’s more notable is being busted because the prompt leaked into the script. It shows how easy it is to stop reading the content you produce with AI before sharing it with others. That’s the real betrayal and is disrespectful to your audience. https://techcrunch.com/2026/08/01/youtuber-hank-green-says-his-ai-usage-is-not-healthy/

I’ve seen many engineers dunking on this position but I think it’s reasonable. Vibe coding is essentially replacing “no code” platforms and arguably the recent sale of Airtable is an indication of how that segment is compressing. Those users have no need for source code, it’s simply unnecessary.

@kevinrothrock@infosec.exchange Depends on your use case but for my needs I’ve found Codex to work best for me at work and ChatGPT for personal usage

Establishment Democrats are only allowed to oppose the GOP on a narrow slice of issues. For instance, only 3 companies have more annual revenue than United Healthcare and they are a major Dem donor. So Dems will never go hard on universal healthcare. This is just coordinating the kayfabe.