What are we doing here?
And: All about Claude
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Happy Friday! Before the news, however, an off-topic note. Tomorrow is the ninth anniversary of the day I quit drinking. For more on how that all went down, head here, but I want to remind everyone here that if you are struggling with booze or any other drug, reach out. If you need someone to listen, I got you. If you need a pep talk, I’m your guy. If you need someone to tell you it’s time to shape up, I’ll try to be that person. Don’t let drinking or some other substance take everything from you. You’re worth more. — Alex
📈 Trending Up: US-EU tariffs? … German-Chinese relations? … US-China trade talks? … back-patting … prisoner exchanges … regional tariffs … Fed independence … corruption …
📉 Trending Down: Media-AI relations … Spain’s cyber rep … budget compromises … China’s home building rate … India’s manufacturing growth? …
Hinge Health’s IPO priced at the top of its range, and closed its first day up about 17%. But as TechCrunch’s Marina Temkin notes, the company is still worth around half of its final private-market price.
What are we doing here?
The White House is working to kneecap Harvard. After the school decided not to give up its independence to the Executive Branch, POTUS and company have worked hard to tear down one of the brightest jewels in American higher education and research.
Not only are the university’s various research contracts with the government being torn up and thrown to the fire, but the White House this week “revoked Harvard University’s ability to enroll international students, delivering a sharp punishment to the elite institution for refusing to bow to the administration’s policy demands,” as CNN put it.
Whether the move is legal is not yet sorted out; FIRE is suing.
If you care about the United States attracting the best of global talent so that they can come and learn and build and lead, this is awful news. One way that America wins is by being a strong magnet for talent. Folks born elsewhere come, found and build companies here in the United States, and often rise to the CEO role. (See: Microsoft, Google, etc.)
One of our chief advantages as a nation is that we’ve long welcomed smart and hardworking folks to our shores. One way we do so is by allowing the world’s smartest kids to come learn at our universities. And then stay. We’re a long way from Trump promising green cards to all graduates from abroad, aren’t we?
All about Claude
You can tell it’s 2025 when we have seen several new state-of-the-art AI models released from rival companies inside of a single week. Following Google’s very well-received I/O slate of announcements, American AI foundation model company Anthropic dropped several new models:
Claude 4 Opus: What Anthropic calls its most powerful model to date, and “the best coding model in the world.” It’s not cheap to use, but can do quite a lot. Per Wired, one customer “deployed Claude Opus 4 to code autonomously for close to seven hours on a complicated open-source project.”
Claude 4 Sonnet: Cheaper to run, Anthropic nevertheless writes that the new Sonnet model “soars in agentic scenarios and will introduce it as the model powering the new coding agent in GitHub Copilot.” That last sentence is a huge endorsement of what Anthropic built, and a huge snug to OpenAI. (You can’t get that mad at Nadella; OpenAI’s Cortex is competing with GitHub’s own coding tools, so all’s fair in love and AI.)
The models were only part of the deluge of updates from Anthropic — similar to Google’s press release deluge, keeping up is a challenge — including:
The new Claude models can use tools like Web search, and swap between tool use and reasoning to answer user queries and requests.
The release of Claude Code into general availability, following “extensive positive feedback during [its] research preview.”
And new API features including a file storage service for use in an AI context, and a tool to connect Claude models to “any remote Model Context Protocol (MCP) server without writing client code.”
It’s great fun to watch OpenAI, Alphabet, Anthropic, xAI, Meta, and Microsoft fight over who can build the best AI model. After all, what’s bleeding-edge today is cheap in a few quarters time, so no matter the initial sticker price, we’re all going to get smarter, faster, more flexible, and capable AI technology in the coming moths.
Hell yeah.
What are those AI safety levels everyone keeps talking about?
More powerful models engender more concern about misuse. Not only in the rogue AI starts writing and executing its own code, takes over the world sense, but also in a more mundane could help a human kill a lot of other humans context. Anthropic’s Jan Leike wrote that:
He’s referring to Anthropic’s decision to activate “AI Safety Level 3 (ASL-3) Deployment and Security Standards,” which means “increased internal security measures that make it harder to steal model weights, while the corresponding Deployment Standard covers a narrowly targeted set of deployment measures designed to limit the risk of Claude being misused specifically for the development or acquisition of chemical, biological, radiological, and nuclear (CBRN) weapons.”
On its face, that’s reasonable. We don’t want AI teaching unwell people how to build dangerous weapons.
However, one AI denizen at Anthropic posted tweets that raised eyebrows, as they implied that if Claude 4 thinks you are doing something very bad — faking data in a pharma trial was the given example — it will try to reach out to the press and lock the user out. Those tweets were deleted, but highlight that we’re brushing up against the AI power-AI risk wall in a new, and interesting way.
In a non-deleted thread Bowman discussed Claude’s safety, if you want more.
Apple faces new trade headwinds
This morning POTUS wrote online that if Apple does not build iPhones that it sells in the United States, in the United States and not in “India, or anyplace else,” it will face a “[t]ariff of at least 25%.”
The move by POTUS to single out Apple is notably petty; Tim Cook even paid the requisite bribe donation to the recent inauguration, and it is still taking stick. Apple shares are off 3.3% in pre-market trading as I write to you this morning.
Yes it would be lovely to write an entire newsletter without mentioning the White House, but as the current administration feels confident making business-level decisions for individual players in corporate America, and especially technology shops, well, here we are.
Apple cannot scale up manufacturing of iPhones here in the United States without a huge amount of lead time. Supply chains, facilities, training, you name it — building iPhones here in the States would be an expensive and iffy proposition. So, Apple is now faced with a potentially lose-lose situation:
Try to move some production of iPhone to the United States, at huge cost and headache so that it can manufacture more expensive (read: less competitive) phones for domestic sale.
Do nothing, and pay a 25% fee to the Treasury, rendering American consumers poorer, the company poorer, and one of the domestic tech giants hamstrung as it faces torrid global competition.
That’s lose-lose.
Ticking trade barriers
Recall that the pause in ‘reciprocal’ tariffs is still ticking (July 9). Don’t forget that tariffs on Chinese exports to the States are still at very, very high levels. And as linked in today’s Trending Up, POTUS is considering at least a 50% tariff rate on the EU, he said this morning.
That’s a lot of trade disruption that we have yet to fully digest as an economy. Cross the above problems with investors betting that we don’t see a rate cut until September, and the economy could get a bit choppy in the coming months. (CNBC has a somewhat worrying trade survey here, if you want to feel even worse.)
Those higher interest rates are probably impacting technology employment
If you have a developer in your life who is gig-less, and finds the current job market to be less than welcoming, I’ve got bad news. As noted above, interest rates are not expected to fall for some time. And that’s probably impacting developer hiring rates, as much as — if not more — than AI today.
Apollo’s chief economist Torsten Slok sent out a chart this morning that makes a pretty compelling argument:
What’s notable about the shift apart from the steepness of the change, is how stable big tech employment growth was for so long. These companies used to create lots of new, high-paying jobs. Today? Not so much.
For startups, the chart implies a more appealing talent market for the hiring side of the table. If those startups want more staff, while mostly I hear from founders wanting to keep their teams incredibly small. Side projects to land a big tech job may simply become projects because big tech isn’t hiring.