spacephysics

FDIC insures deposits up to 250k. Big institutions basically all have this government backing. If an incident occurs, FDIC has deposits available next business day in event of a bank failure

I think he’s the exception that proves the rule. A truly rare, exceptional programmer who not only has the:

- technical skills to build an engine, language, and game

- interpersonal + communication skills to build, manage, and lead a team

- business acumen to take the above 2 skills and create profitable businesses (plural emphasized)

Not trying to lay the idolization on too thick, but an average developer having any one of the above would likely be a multiplier in their career.

One annoying thing having a guard model is ensuring its FP rate is low enough to not be a headache, and its FN rate low enough to be worthwhile to use. Likely also means you’ll need to either use a frontier-close model, further increasing costs as well as round-trip time.

Or maybe some system 1 model?

For my setup, i have claude code in its own VM, whose VLAN is heavily restricted outside of the internet, my forgejo instance, infisical for secrets, and one way SSH from my machine to the VM.

All secrets are shared via infisical secrets store, so the actual secrets are never shown to the agent. Dev mode can be used for running the system locally during development, and those secrets may be seen in some cases. The foregjo user the agent has access to can only create issues and PRs. Manual merge required, after which a CI pipeline handles the packaging and “production” deployment of the app to my app server VM.

I think a setup like that (or discouraging these agents access to the same computer people keep their finances/health data etc) should be discussed/championed more.

My work encourages just running the harness on their system (which I get for practical reasons among others), to which I then put my work laptop on its own VLAN.

Overkill, for sure. But to me the risk of personal data exfiltration is far too high with these vibe coded harnesses and models which are more and more opaque.

Im all for fighting for digital privacy rights. But referencing the purge with a smile begs the question, are you yearning for a purge-like scenario? Some fantastical idea that you’ll be the main character in an anarchic chaos?

I think there have been a positive number of cases gone in favor of digital privacy recently. This being one of them. It’s unfortunate, but it’ll just take time for these manners to go through the courts.

I’m quite naive regarding the details of the legal system, wonder why we can’t expedite certain cases when a SCOTUS precedent has been established, or some form of expedited legal notice. We shouldn’t have to wait for a perfect case of a flock violation, then wait for someone to be brave enough to spend the money (or if pro bono just their time/reputation) to file a case and go through the system for the next x years

This rhetoric of violence and rebellion is wild on a tech forum

Why the downvotes? Genuinely curious, I do this and don’t know of a downside after setup (with help from {AGENT})

The second term I think was largely handed to him given the Democrats had no primary for Kamala, and frankly were tone deaf with what the “other side”’s needs were in a candidate.

Politicians flip flop, or hopefully truly develop different opinions after experience/learning more.

Kamala came off as a piece of clay molded in the last minute to appease groups just for their vote. A good politician makes that “molding” process come off more naturally. Her and Walz did not really resonate with voters, they were just not Him.

And that is a lesson I hope democrats or a third party has learned that they can’t just reflectively say “I’m not Trump,” they need to show that through their principals, track record, and message of unity.

Hillary admonished trump voters, many of whom are middle class union/laborers. Biden claimed he was pro union then broke up the railroad strike. Kamala couldn’t hold a conversation without some sound bite failure (doesn’t make her bad, just not a politician) and wasn’t elected in any primary.

I think we need to better hear and learn the context of the other side’s perspective to understand why they voted the way they did, then provide a better option not just a different one

You assume those voters believe Jan 6 was an insurrection. Many of them don’t. Whether thats a fault of the media/bubble they’re in or gross naivety/negligence is a different discussion.

If you go with the presumption Jan 6 was a mob that got out of hand (not my position, just one of the perspectives), then the idea that he wasn’t able to enact his agenda due to the opposing party, coupled with the shenanigans of the other option (no democrat primary for Kamala being one, with democrat party having done similar in the past with Bernie), then claims that the democrats are spinning Jan 6 and risking another term with him seem more reasonable.

Whether I or you agree with those premise is separate from what the zeitgeist is for much of America

In a democracy where everyone has the right to vote, I think the “downside” that must be accepted is there will exist a spectrum of voters. Voters who only listen to sound bites, voters who mindlessly vote tribally, to voters who research, voters who understand second and third order effects. We also can’t assume we have total information of how the elected official will act throughout their term.

Even something as simple as, attractiveness mattering more in presidential candidates since the advent of television in homes.

At the same time, we had a system whose establishment largely looked down on middle class Americans who didn't go to college. Early to mid 2000’s through the great recession saw the middle class tighten, with no real improvement to healthcare or affordability (attempts were made but ultimately medical costs are still rising). Many middle class people rely on things like cheap gas, child care, and affordable heating among others.

Then you have a system whose leaders actively works to increase gas prices (regardless of EV incentives, you’re asking individuals to purchase new cars when a difference of 1 dollar in gas goes a long way), some states subsidizing green energy programs by additional taxes on oil and natural gas heating costs (just switch to heat pumps and take on a 0 interest loan for 10 years), and rising childcare costs partly due to regulations in childcare businesses.

Then you have someone who claims to be an outsider (he wasn’t), who clowns on the existing elites in debates, and makes vast promises. The other party doesn’t say anything about positive change they’ll bring, just that they aren’t Him. Why not try something different than the status quo over the last 2 decades?

With gas prices up, the Epstein scandal still unresolved, and another ME war, i think if many of middle America were told these would be the outcomes they would not have voted him in again.

We should have more empathy for our fellow Americans, and try to understand each other instead of assuming they want the worst for your side.

Likely easier said then done, but why not separate infra out from shared vs dedicated hosts? Something like, actions will still run but at some degraded capacity, where actions and PRs are still at least functional.

My company has heavy handily moved from self-hosted version control system to github and now we’re suffering more outages. Though the self hosted system had less features, and we needed to maintain our own jenkins infrastructure for CI/CD, at least we really only suffered from very major cloud-provider level outages

I agree with this framing, but still trying to reconcile staying ahead of the curve without adopting full hype train or doomer mentality.

How are you keeping your “human” addition to the loop valuable, is it through the time spent on the software craftsman hobby?

Im in my early 30’s, and trying to stay ahead. I feel it was easier prior to LLM’s, and now its tough to even know where to focus skill building.

Live free or die

This is a bit of a straw man “do the ends justify the means” statement, with hints of anarchism read between the lines.

Western democracies haven’t really prioritized “war on drugs” (for lack of a better term) over the last probably 10ish years. Yeah there’s still funding and initiatives and some lip service.

Instead the focus has been largely infighting and major political factions fighting for political power in the next generation (“tech bros” political class for one)

So it’s no surprise state-sponsored (and just demand driven) drug trafficking into the US hasn’t received much action.

There is some hope though. With rare earth mineral access being a major national security priority, and many cartels controlling those areas in SA, the converging interests may produce some net positive. Background: https://youtu.be/q_n-74ifBZs?is=nw5g-mmPmKphGH2W

Though in general I agree, i think in HN world this seems more or less old news/obvious, while mainstream still doesn’t know (or maybe care to understand) thats what likely is going on.

I had a friend reach out in a slight panic after the Anthropic employee publically left, asking sole quite apocalyptic questions around AI given the sandbox “break out” and the Anthropic affiliated employee(s) saying theres something like > 10% chance AI kills all humans in the next decade.

This fear mongering needs to be exposed, so we can (hopefully) have more measured discussions around next steps. Stop all progress for regulatory capture? I’d rather not. Hold companies who commit crimes like hacking huggingface or Ruby dependencies liable (not to mention all the copyright issues)? Seems more measured to me. But i digress.

I think the world events we’ve seen over the last 4 or so years (especially last 12 months), will mean both new (and existing) data centers and refineries will begin to have counter measures beyond a security guard and some fences.

Anti-drone equipment for one. The pipelines themselves are tough to secure given how long some are.

Closer and closer to a cyberpunk world with faux (or real) military at these data centers and refineries in areas that may not be in direct active conflict.

Imagine building a new refinery, the money and time alone, only for a few thousand dollar drone to be snuck in and launched? Insurance on these facilities going to start requiring far more elaborate physical security measures.

Would be a great way for up and coming researchers to learn and apply skills before attempting to work in labs pursuing novel research.

Maybe some joint effort across a large number of research schools which function as “proportional to number of students in your program, X number of validation studies must be performed” with maybe some monetary or access incentives aligned.

Since we’ve had student loans all but guaranteed, and therefore cost of schooling skyrocketing, the flywheel of pursing higher profits and more bureaucratic cruft (looking at ratios of university staff to students), its no wonder that incentive has extended to research.

Also, the $20/month subscriptions are HEAVILY subsidized, so it's not an apples-to-apples comparison really

We should be mindful of the context that many of these providers VERY likely have been selling their subscriptions at a substantial loss

So as much as i agree “more profits to stakeholders screw the customer”, i think its more of an emergency to get to profitability before the music stops.

Money market/treasuries (even ETF like SGOV) gets pretty close to 5% when typical savings rate is a bit under

If treasuries “fail” we have a different class of problem.

Ah yes, California the land of regulations. A compliance tax for the tire companies, less freedom of choice from consumers, and likely inferior products for a hefty subset of use cases. Tale as old as time.

Most of my model usage comes from my work’s model selection (which is now down to just Claude models)

I’ll try out the latest models, but mainly stick with Claude only because I’m most used to its quirks and how to work around them. I imagine this is part of these hyperscalers playbook.

I will say though, I miss Sol model at work. It with Codex was amazing at first-shot understanding. Claude i need to scope out where to look otherwise a large portion of my token budget is eaten up

This is a national security nightmare. For better or worse, fossil fuels have massive energy densities, which makes economics in defense context (and others) extremely favorable.

Even with today’s energy polices, electricity costs have skyrocketed (this was the case before the data center demand). Couple that with electrification of vehicles, phase out of natural gas/oil heaters, I think its reckless abandon to push policies which harm middle and lower class families. Not to say some of these countries even consider nuclear energy “evil”. Absolutely ridiculous.

I experienced similar reaction when i was doing keto. It became far more difficult to burn that summer

Tell that to struggling family’s electric bills

I imagine part of the performance issue is their encryption flow? Their search is sub par even with the on-device search enabled, but besides that i’ve been a happy customer for a few years now. Catch all domain, multiple domains, their cli lets you download all messages and I setup a RAG flow to better search.

Regarding outages, this is the first one I’ve actually noticed and affected me. Obviously not great, but maybe I’ve been deluded in seeing GitHub‘s fiasco of what acceptable means.

You can just make the tool calls restricted/scoped to whatever the calling account has access to (or in this case the repo)

That way even if the LLM broke out of the system prompt the worst case would be similar to a 404 or 401.

Why are we giving these processes super user access? No reason to have the executing loop/chat turns/tool calls be scoped to anything but the narrowest permissions.

If the agent truly needs data/permutations across different accounts or repos, treat the tool calls like any other API that needs to do that kind of work pre-LLM

I think this seems purposeful, as there's 2 opposing forces at play: - Have a model that follows the users instructions - Have a model that follows the system prompt instructions more

For the 'safety' argument (Re: Fable), they need these models to have basically a 2-tier instruction system, but given LLMs aren't great with actual Logic unless they program it out to test, this runs afoul and we get one or the other.

Feels like optimizing for either precision or recall, but can't have both

Accountants can be expensive, especially if your books are messy, or have poor accounting practices from the start.

Systems like quickbooks, hubspot, payment processors all have tiers where yes on paper they make it easy to properly setup good accounting practices, but you’ll spend an additional 500/month+ to get those features.

Hiring an accountant to clean up the books and do quarterly book keeping is equally as expensive if not more.

Especially for small service based businesses, where margins can be tight, revenue can fluctuate heavily MoM, committing an additional $6k+ per year just to keep books organized is non-trivial.

As an experiment, i gave all our finance data for 2025 to an agent, and it did quite well after spot checking. There may be a middle ground where users can do exports, verify with “real” software, and have agents handle contextual classification to considerably cut down costs

I’d slightly disagree, the profile of people who go missing is as important as a random chance there is a coincidence. Former military officers, high-level scientists. These individuals have training, money, and live in areas where this tends not to happen.

A disappearance of someone from the above background, vs someone who is say in midwest rural America or near areas where human trafficking crimes occur at a higher rate than normal, matters.

Further, their research/knowledge of sensitive government material also implies they likely have some form of overwatch or at least minimal monitoring for foreign agent threats from our government (or had in the past). Its not uncommon for high ranking military officials to have some form of training in counter surveillance tradecraft for this exact reason.

The odds these events are due to a foreign adversary given the multiple wars and geopolitical tensions are not negligible