With modern automation and AI, tracking and adjusting paths is better every year. Also, anything with malfunctioning movement will quickly descend and burn up in the atmosphere at that very low orbit.
There are many biotech startups and private research labs thriving and paying high salaries with excellent benefits for that specialty right now - focused on genetic testing, editing, and longevity. Before moving abroad, widening the search outside of academia and considering moving internally might be worthwhile.
Correct, everyone complains about "enshittification" but they fail to recognize that the most egregious cases occur when a company with a dedicated founder/leader transitions to rule by committee.
Venice (uncensored privacy AI API and app co) took a year to expand their self-hosted model selection to routing hundreds of other models. It's harder than it looks to get customers. But they did grow to 3M users and >$50M ARR as of a few weeks ago. So go for it if you've found an easy way to do it.
If you want to avoid bans, Venice is another good option since their focus is uncensored and privacy. They run models themselves alongside offering OpenRouter-style routing for frontier and niche models - but at least they fully anonymize the user and never ban.
Or you can get mad at journalist-bloggers for writing unscientific selective fearmongering hit pieces catering to culture-war-polarized confirmation biases.
His operation tangibly and physically contributes more to the material advancement of humanity than most of the armchair critics in the comments do - I think that matters more for character allegations. And this first of its kind refinery in America is far better than the alternative of continuing to rely on far dirtier refineries in other countries.
FYI lithium is fine at those levels, and is even biologically beneficial, commonly taken as an OTC daily supplement at much higher doses.
They got privileged early access to an unreleased frontier model to harden their systems, with Anthropic engineer support, and likely were able to use it to make other product optimizations tangential to security too. A blog article afterwards is a cheap price for unlocking that access, regardless of how well it paid off.
That's why they added a verifiable E2EE mode that encrypts before leaving your device all the way to the GPU's TEE. You can proxy and see the shape of the request if you like. The platform supports no-KYC signups too, so if you care you can disguise your user too.
When using their platform via web/app there's a temporary chat option that avoids local browser storage entirely. You can also wipe the local browser storage whenever you want.
Yes, local for anything that can run locally. For higher-end model needs there are privacy platforms like Venice (https://venice.ai/privacy) with ZDR legal contracts and multiple E2EE options for their open-weight models. The OpenAI/Anthropic/Google models are also available through through them but at least your identity is anonymized, though the contents of your prompt could still be stored by the destination company.
That would be an improvement in the same spirit, as the modern CRAs are also unconstitutionally limiting over what one chooses to do with their own property. The scope should be limited to government-involved services and facilities, as those must serve all possible taxpayers. We live in a more connected, option-saturated information age where even bigots more often than not understand the utility of at least doing business politely and making nuanced exceptions. Egregious offenses are corrected by social pressure and business competition. The current regime of ambulance-chasing liabilities inserted into every organizational, contracting, and hiring process harms far more people of every race and sex than it helps.
The "pen-testing" discoveries go both ways. In Iran, Chinese HQ-9B surface-to-air missile systems and YLC-8B anti-stealth radars failed to intercept any aircraft. In Venezuela, Chinese JY-27A early warning radars failed to detect approximately 150 incoming U.S. aircraft. In Pakistan, Chinese HQ-9B and HQ-16 systems failed to intercept Indian strikes.
Most forms of company civic greatness in the past were essentially pledges, much of the time unspoken. It's certainly possible, we don't need to be cynical.
If you are planning to get a Tesla car with PowerShare (it's slowly expanding to the Model Y and other vehicles) then you only really need one Powerwall 3, because the car when charged acts as ~7+ powerwalls worth of backup.
Unfortunately the MCP for external agent tools is flawed at the moment, it returns a format differing from its stated schema. So it doesn't work with OpenCode.