smallerfish

Location: Costa Rica, Miami

Remote: Yes Willing to relocate: No

Technologies: Any web stack

Résumé/CV: https://reuben.firm.in

Email: rf@4rc.io

I specialize in Fractional CISO and Fractional CTO engagements. I'll get you through SOC-2 or ISO-27001, hire your team, prototype your product, or help you with anything related. 27 years in startups and small tech companies, with 18 years in technical management and exec roles.

Everybody in tech should read Careless People, not just to learn details about Meta. I'm sure many of us (myself included) have worked at places where the technical challenge outweighs any concerns about the overall company's mission or effect on the world at large. In Facebook's case they have a lot of blood on their hands.

> A human is just a few chemical reactions, and fairly stable ones at that.

And humans are controllable. Pump the system full of lithium and morphine, and your human becomes much more docile. You don't need to understand the full system in order to constrain it.

Fair, but is it possible you've tried it on a setup that wasn't done particularly well?

Fair - I used one only briefly when they first came out. Are there any serious attempts on desktop that I'm overlooking?

I think he's bang on with his observation that overlapping windows are a relic.

My wife, whose technical level leans her towards wanting to replace her mac mini when a VPN that she'd installed and then disabled left hostile firewall rules in place ("the internet is flaky on this computer"), is jealous of the tiling setup on my desktop (was hyprland; now sway with a custom dwindle layout). I think tiling has an audience beyond nerds, but I think nobody (perhaps with the exception of omarchy) has made a serious attempt at packaging it in a way that's accessible to business/productivity users. KDE's tiling attempts were, at least last time that I checked, somewhat bolted on and half-assed. I wish omarchy success in de-vibing their codebase, but it would be great to have some serious competition in the space.

I stopped using them last year. They were too late to see the AI juggernaut coming and their attempts at responding have been underwhelming. They should have been pouring resources into Fleet from 2022 onwards to make this, but instead it's been a series of missteps. This looks like approximately the right offering, but where's the moat? It seems like just another Cursor/Antigravity.

What are they examples of? Opus 4.8 was much better than the infamous 5, and I find Astra generally competent.

On my home PC I use restic, and (shockingly) I run it manually (via a script). I have a terminal banner that nags me when backup hasn't run in 24 hours, or displays spanish verb tables when it has (so I will notice if the banner doesn't show). I've been burned in the past losing a backup cron, and dead man's snitch seemed overkill for my needs here. And I've also forgotten what backup system was in place on a prior install, and it took me way longer than I liked figuring out how it was setup when I needed to do a restore.

Obviously for production systems automation & monitoring is the way. I kind of like this setup on my own machine though.

Tasking AI with "solving climate change" is exactly a scenario in which the doomer worry comes true.

> Nobody is doing this.

https://www.axios.com/2026/09/10/ai-investors-anthropic-doom

> In part because you can't simultaneously argue that something "is not going to happen" but also will happen if we regulate it.

Even if the US regulates it (not going to happen: cf Trump administration), China won't, because it's a competitive advantage against the US to have better AI.

> It took 88 hours and who knows how many kwh to solve a problem that no other human has solved, and Humans have already solved breaching and destroying systems that can't be accessed over the internet (see Stuxnet).

> that can outsmart

Just to reframe slightly: out-remember and out-persist, not outsmart. As far as I understand, the math problems have been cracked by combining stuff that humans originally developed, but hadn't been able to put together yet. We still have no evidence of AGI, nor is it guaranteed that LLMs or diffusion gets us there.

The nature of the problem is not infinite. It's solvable with some manageable coordination. It is definitely urgent. It is very unfortunate that we have clowns running the US right now, but other countries might coordinate themselves much better.

Fair. That doesn't mean we should throw our hands in the air. The genie is out of the bottle, we know the nature of the threat, and we've spent 25+ years figuring out how to make distributed systems theoretically secure -- instead of trying to "slow AI research down" (which, again, is not going to happen) we could instead be yelling about NOW being the time to secure the shit out of your essential systems, train your staff, etc. Yes, an uphill battle, but more productive than trying to stop the train that's already left the station.

Isn't the "kill us all" concern met by keeping nukes & essential infrastructure airgapped? This is something that should be the case already, and is certainly urgent if not the case. Instead of doomerism and calls to slow down the development of AI (not gonna happen), the conversation could be pivoted towards thorough threat analysis / audit.

Granted that a rogue adversarial AI in a sci-fi scenario could theoretically go into paperclip mode, but if "we" maintain control of energy generation and civilization threatening weaponary, the damage is at least limited. AI needs a lot of power, and AI can only (theoretically) breach & control what it can access through the internet. Even the scenario of nation-states deploying adversarial AIs against each other is just an extension of this and current deterrence / realpolitik.

And, it's not like the scenario of all models teaming up with each other to squash the human plague is plausible. We have AI with which we can protect ourselves against the threat of rogue AI.

We do use a drone when we know of a site where deforestation has taken place. The gap we're trying to fill is early detection of forest intervention, so that we can alert appropriate agencies (who are generally under-resourced) and get an inspection to take place while activity is ongoing.

We're volunteer driven and currently rely on personal donations, so we're scrappy. We could probably justify somewhere in the region of $10k/yr. The jets/balloon angle is interesting, thanks - hadn't occurred to me, and I'll search through what's out there.

I wish Planet had better pricing for nonprofits. I co-run a conservation nonprofit focused on deforestation in Latin America, and our affordable choices are between 2 year old Google Earth imagery, 10m resolution Nimbo imagery (ok, but not sufficiently detailed to be evidentiary), and Sentinel 1 SAR data, which is useful but very noisy. Planet quoted us something like $30k per year for a strip of coastline that represents something like 5% of the territory we're monitoring, so it's just not a viable option under their current pricing model.

Yeah and there's two ways to get there - planned and unplanned. It'll happen one way or the other.

We need sufficiently severe penalties for vibe coding. First time out we break your legs, but if you come back with another vibed patch its around the back of the barn for you.

> We’ve never had a technological revolution that didn’t yield more jobs

a) How many technological revolutions do you count?

b) Of these, which directly created jobs vs in which did the job numbers eventually recover?

An LLM that could mutate its own weights might be closer to it. The various labs must have experimented with this, but it seems that it would be economically prohibitive to productize.

I did the same this week.

The speed is the first big contrast; I have a routine multi-step skill that I run several of per week. Opus 5 was routinely taking 2 hours to do it, while older Claude models took around 20 mins; Codex restored that speed.

Second is legibility. Somebody wrote in one of the related discussions yesterday that Claude's current linguistic contortions could legitimately be considered damaging to mental health, which doesn't seem (too) hyperbolic to me. Codex (Sol) isn't perfect but it's much more direct. And so far I haven't seen it display much of an attitude, vs Opus's infuriating passive aggressive sulky know it all personality.

I slightly prefer Anthropic to OpenAI as a company, but I will vote with my wallet and discontinue my max subscription unless Anthropic does some serious damage control within the next week or two.

I really wish these agents attem[pted less personality and were more mechanical. I find Opus 5 to be incredibly annoying. It's also significantly slower - repetitive tasks that used to take 10-15 mins now take around 40 mins. And, its writing is much worse, as stated elsewhere in the thread.

Having it write the initial draft can be educational if you read through as an editor and research and/or ask questions on everything you don't understand (and also work to eliminate slop construction and tics). Certainly not the same as doing the initial draft yourself, but there's still an active process to it.

I am 100% for creating software that you find interesting with LLMs, but you need to do better product marketing to get anybody interested in trying it out. Two things:

a) Write the copy yourself. "hinge", "saying it plainly", etc. Nobody wants to read unedited LLM text on a website.

b) Make a video of you using the software. Show why it's interesting.

Actually they did something like slash and burn agriculture, but they apparently rotated sites, allowing forest regeneration.

I've been hacking on my own task manager on-and-off for years (https://the.do.zone). It has no database; the primary store is the browser, and you can optionally sync to local file system (if using chrome), dropbox, or google drive (less optimal for my needs, because their js oauth implementation is very conservative about token length.) It has some crude but functional diff resolution, so that you can sync across multiple machines. Browsers _could_ provide better tooling to make this kind of app architecture more convenient to write; and theoretically you could imagine cloud providers providing some kind of standard for datastores. But there's no market for it; non techies really don't care about the location of their data to any great extent.

> but without any lived experience perhaps that's an unreasonable expectation.

I don't think that follows. We learn via our perception (i.e. our inputs from the outside world); our perception is mediated by some kind of filter, and memories get written or altered. An LLM on the other hand has static weights. You can provide an LLM a "facts.md" (or whatever) file that "teaches" it stuff that's not in its weights (via the context), and it can use the combination to produce something, but that set of information never modifies its core weights. You could imagine a robot with an LLM inside it that has all kinds of sensors, and processes that write the observations from those sensors into a bunch of databases which are RAG'ed into prompts to simulate learning and experiences, perhaps (jefe likes her coffee with almond milk). But my point is that, architecturally, they don't learn - and that's the bigger obstacle.

PS garcinias seem to be generally delicious.

I wrote a framework for my own use that uses Kotlin + HTMX - https://github.com/reubenfirmin/zoned. The goal was to see if I could create webapps that were fully typed, end to end. It uses Kotlinx.html, which provides a jsx-ish dsl for writing html.

I did the first 90% by hand, and have done the last 10% (and README) with Claude, just to get it out there.

The voting system could be enough if downvoting was added.

AI writing is not the problem - low effort is the problem. Low effort AI articles are full of tics which are obvious, if you've done a lot of AI writing. To write well with AI you need to spend a good deal of time editing.

If you submit something that's low effort but has a clickbait headline that appeals to HN, you may well make the front page even if the article is lightweight (it does happen!) This is true both for AI and for human written articles.

On the flip side, somebody could spend an enormous amount of effort creating a masterpiece with AI. Penalizing that because of the tool that was used is arbitrary.