HNHacker News
TopNewBestAskShowJobs

hammyhavoc

3,414 karma · joined February 8, 2018

CEO of Split An Atom and Previous Magazine. Composer for film and games including All Walls Must Fall (Nintendo Switch and Steam), Manta, and a smattering of TV-syndicated and theatrically-released nature documentaries. Makes music as The Orion Correlation. Dev. Cypherpunk.

Autistic and ADHD. Not terribly socially apt, likely sorry if we are having a disagreement in the comments and would buy you a drink.

submissionscomments
hammyhavoc··on I tried vibe coding in BASIC and it didn't go well
Fantasy.
hammyhavoc··on I tried vibe coding in BASIC and it didn't go well
They'll never be fit for purpose. They're a technological dead-end for anything like what people are usually throwing them at, IMO.
hammyhavoc··on Giving Up on Element and Matrix.org
The big platforms always use a variety of solutions. Always some kind of database of hashes of known CSAM available via an API. Always human beings on a trust and safety team. Takes considerable time, money, and emotional toll.

Not just anyone can access all of the SaaS databases and it's approved on a case-by-case basis. Obviously not all CSAM is in these databases, and gen AI output adds another hurdle to preventing this kind of malicious behaviour.

Personally, I feel a potential solution could be that user clients could opt into referencing hashes of media before it is downloaded, and that could either be handled by their home server, or a third-party. However, something like that is really idealism and not realistic at this point, despite being a great solution in a perfect world where people maintain encryption, yet don't expose themselves to CSAM inadvertently.

In terms of the Matrix.org homeserver, perhaps an alternative (though would prove unpopular) is only allowing Matrix.org users to upload media if they're a paying supporter of the Matrix.org homeserver with a valid payment method on their account, thus obviously linking their identity to their account.

https://matrix.org/membership/

The knock-on effect would be that it may offload some users onto their own infrastructure and alleviate the moneypit of a public instance, or directly fund the public instance itself if they do choose to support it financially. The Matrix.org instance doesn't support users adding bridges, so it isn't that much of a stretch to forbid media uploads from non-supporters. Portable identities would make this a no-brainer, but even without them, I still feel as if the public instance is largely an incredibly generous demo to help folks take Matrix for a spin. Decentralisation is such a major selling point that's vastly underutilized.

I do think that people should at least attempt to become decentralized and manage their own community by themselves, be it public or private. Large public homeservers are always going to be considerable targets for this kind of problem. It's a lot of digital eggs in one digital basket for a lot of reasons.

Nothing stopping people uploading images onto a third-party file storage system and providing the URL if they absolutely need to share a screenshot or similar on the public Matrix.org instance. Anyway, fleeting thoughts.

hammyhavoc··on Giving Up on Element and Matrix.org
Can you quantify this "better"?
hammyhavoc··on Giving Up on Element and Matrix.org
Not to mention the great "This Week in Matrix" RSS feed they do that I've been enjoying for years.

https://matrix.org/category/this-week-in-matrix/

Incredibly supportive of anything within the Matrix ecosystem.

hammyhavoc··on Giving Up on Element and Matrix.org
Element X is (e)xcellent. Just wanted to share that.

Been using Matrix via Synapse since ~2019 and all the Matrix odds and ends require the least maintenance out of anything in the stack. Any uptime issues I've experienced have been my own silly fault with incorrect configurations.

Even Element as it stands without the X builds is leaps and bounds ahead of anything else I've tried (and I have and continue to try a lot of FOSS, but nothing does what Element does as a client, nor what Matrix does as a protocol).

hammyhavoc··on Giving Up on Element and Matrix.org
You're thinking of Element: https://element.io/case-studies/nato

Now do Nextcloud and many other projects.

It's FOSS. There are volunteer contributors.

https://matrix.org/blog/2025/06/dispelling-myths/

> Matrix is not just an open standard for secure communication, it’s an openly governed and collaboratively developed ecosystem of projects powered by a growing community of volunteers and vendors. In this way, Matrix exemplifies the open source ethos, encourages greater innovation, and defies those who would try to build businesses based on extractive behavior and vendor lock-in.

Oh look, more calls for volunteers, despite "running on European taxpayer money": https://conference.matrix.org/volunteer/

Sorry, but idiotic remarks like this stink of having an agenda.

You don't even put a name and identity to your HN account and provide zero sources. At best, you're clueless, at worst you're spreading FUD.

hammyhavoc··on Giving Up on Element and Matrix.org
Exactly.
hammyhavoc··on Giving Up on Element and Matrix.org
"Pull requests welcome" is all I can say to anyone who puts energy into long rants about the "state" of a FOSS project.

If you think volunteer dev is so easy then get stuck in and be the change you want to see.

hammyhavoc··on The new Beeper – mostly on-device withe full E2EE
https://engineering.fb.com/2024/03/06/security/whatsapp-mess...
hammyhavoc··on The new Beeper – mostly on-device withe full E2EE
Never have issues with the likes of Meta when self-hosting bridges for Matrix. Beeper's centralisation is likely the issue.
hammyhavoc··on Signs of autism could be encoded in the way you walk
People have always remarked about my gait, so this is definitely a "no shit" kind of thing lmao.
hammyhavoc··on Mark Cuban: Why AI Will Create More Jobs, Not Fewer
I don't think it matters whether he's right or wrong when he's not good at making tech-adjacent predictions, period. He's only still put on a pedestal because he's a billionaire, but is mostly irrelevant to any conversation of 2025.

Anybody amassing billions of dollars in personal wealth has a non-desirable character flaw and is a potential danger to the human race. You know what would create jobs? Circulating wealth instead of trying to pump-and-dump crypto shit and AI vaporware.

That we value the thoughts of billionaires with zero relevance to the conversation at hand is exactly what is wrong with this. Mark Cuban having any thoughts on AI is celeb gossip. It's like Justin Bieber telling you that you're not clocking the grass is green and the sky doesn't exist—who cares? Total brainrot.

The billionaire-class who will never be affected by the things they comment on from their ivory towers are visibly uncomfortable on a level that multiple billions of dollars does not comfort. These are the people building underground lairs in Hawaii, armoured weenie wagons with laughable bulletproof window stage demos, and the same people who will tell you that you can empower yourself by buying the thing that's going to make them wealthier.

Mark Cuban is really saying "when those with less than me decide to each the rich, please don't eat me for investing in this thing that I told you wouldn't hurt a bit, I tried to comfort you. I'm not a bad dude. You want to be wealthy and powerful too, right?"

If it was someone more reputable with a knack for making predictions in this specific socioeconomic capacity regarding new technologies then they are perhaps more worthwhile to value the opinion of—when considered alongside many other experts and what they say, but not individually.

People shouldn't ask me about volcanoes because I'm not a volcanologist, so we shouldn't ask billionaires with very few relevant skills and not much in the way of relevant thoughts to predict what will happen. It's like being a serf and asking a king if it's all going to be okay for the peasants—of course they will say it will be.

Mark Zuckerberg was adamant that the Metaverse was going to be the next big thing and burned billions on it. Even renamed Facebook to Meta. Zuck and Cuban are saying AI is going to `xyz`, so I'd take what either of them say with a colossal pinch of salt.

Personally? I think life will largely carry on as it was. Like crypto, the niche becomes increasingly full of snake oil, vaporware, conmen et al, the public eventually starts treating the buzzword negatively, and the hype cycle begins all over again with something else. Most of what people are throwing LLMs at is inappropriate.

I'm personally more interested in machine vision and machine learning. I personally hate the term "AI" being used to refer to LLMs and generative AI. AI is a big field and all the interest is on what's basically a gimmick to me. The term is now so poisoned that it's like people believing anyone who works with cryptography has something to do with blockchain grifting. Likewise, people immediately assume decentralization must refer to blockchain. It's been hugely unhelpful with little in the way of meaningful technological breakthroughs.

Okay, LLMs. Cool. What's the value here? They're shit for customer support. They're shit for writing new code that hasn't been done before, and still pretty shit for things that no-code solutions could knock out far more reliably. Their output is trite drivel in terms of fiction. Their output is diminishing the signal-to-noise ratio.

People like Mark Cuban are worsening the signal-to-noise ratio too. We don't need to hear from Mark Cuban anymore. He's sat on billions of dollars in personal wealth and still wants more, just like Elon Musk, and has a long history of being the loudest and most unjustifiably over-confident dude in the room.

Like with cryptoshit, the "AI" buzzwordery is another king-making grift for the most part. The days of any semblance of good intentions or actual substance are long behind us. We are into the era of "we will reach AGI" borderline-religious bullshiting of ourselves as a species because it is profitable to peddle lies for people to believe in. This is all without actually admitting that what they're calling AGI is not what is classically defined as AGI.

Elon Musk supposedly desires to make an LLM that finds "truth" (Grok), yet it fails spectacularly at that. If he ever did succeed in that pursuit, either he would destroy it or it would destroy him one way or another, figuratively or literally. It can't reason—it's a next word prediction model. To believe otherwise is unscientific and makes someone irrelevant in their thoughts.

When the dust settles, history books will see it for what it is—digital pet rocks and hand puppets. Some people are talented ventriloquists, but you shouldn't trust what the puppet LLMs tell you when their pre-prompting makes them follow a specific agenda. You only need to converse with the average LLM (ab)user on HN who has an LLM summarise an article who argues something non-sequitur or false because their AI summary was so off the mark and usually admit they only skimmed the article (bollocks), or had it summarised.

Personally, I think increasing the scale of tokens in a model leading to worse output means that AGI will not be reached via LLMs. It is a technological dead-end and gimmick, IMO. It's a cool trick, but it's all sizzle and no substance.

If there was truly substance, where is it? Why aren't all the bugs and roadmap of issues sorted for major FOSS projects? Why are companies who implement AI customer services not uncommonly ditching it after a while? And no, companies that offer AI-adjacent products don't count—selling pickaxes in an illusory gold rush, IMO.

Copilot is on GitHub, but the quality of code is increasingly dogshit. Claude Code can read and write repositories directly, but vibe coded slop is now a meme.

hammyhavoc··on I want an iPhone Mini-sized Android phone (2022)
It's interesting to see someone bring up ThinkPads!

I have been saying for a few years now that I want the "classic ThinkPad of phones" with readily available parts, sensible design choices (drainage holes on ThinkPads were a boon! Same with quickly swappable battery handles!) and some kind of afterlife that goes beyond whatever the manufacturer or OS developer decides it is going to be.

To my knowledge of keitei, there aren't any Richard Sapper-ish ThinkPad-styled ones, but you do occasionally see red flourishes like the ThinkPad nurple/TrackPoint, but they're usually on things like a SIM card tray, but not generally contrasted by black, but similar impact to Sapper's distinctive style.

I love me some utilitarian industrial design—I think that's what I'm searching for in modern phones. Sadly, I think that mentality is leaving us as the likes of Sapper pop their clogs and take the mindset with them. It's definitely somewhat oldskool values buried amidst all this.

Sent from my modded classic ThinkPad next to Richard Sapper's '70s Artemide Tizio lamp with an aftermarket LED upgrade to make it energy efficient. :- )

hammyhavoc··on 23andMe is out of bankruptcy and it still hasn’t substantially changed its ways
You're describing the process working—DNA gave investigators a lead, which they followed up with legally obtained, direct evidence tying DeAngelo to the crime. That’s not guilt by association; that’s good policing. Framing that as something sinister ignores the outcome: a serial rapist and murderer finally faced justice. That’s not dystopia—it’s what society should be doing—or would you rather a rapist, kidnapper and murderer walked freely?
hammyhavoc··on Elon Musk Wants to Turn AI into a Cosmic Religion
"Translation? Musk believes that the most successful AIs will be the ones that maximize things that matter to conscious beings; things that feel good, are rewarding, or extend life. In Musk’s view, that means aligning AI systems with long-term human flourishing, not short-term profits."

"MechaHitler"/Grok is not convincing me Musk is someone to listen to when it comes to LLMs.

hammyhavoc··on Elon Musk Wants to Turn AI into a Cosmic Religion
You say that, but marketing seems to be acceptable though.
hammyhavoc··on 23andMe is out of bankruptcy and it still hasn’t substantially changed its ways
DNA isn’t guilt by association. Cops still need real evidence—this isn’t CSI fan fiction.
hammyhavoc··on 23andMe is out of bankruptcy. You should still delete your DNA
What makes you believe this is coming? What evidence points to this inevitability?
hammyhavoc··on I want an iPhone Mini-sized Android phone (2022)
I'm definitely open to the idea of foldables or even flip phones (perhaps even enthused!). I'm gutted that the Japanese "Galapagos syndrome" keitei are becoming extinct with fewer and fewer releases each year. The ones that are newly available tend to run Android 10 (yikes). The keitei were always very tasteful, ergonomic, and sensible. Sure, not always flashy in specs, but they didn't need to be when they prioritized the form above everything. Would love for the rest of the world to pick up this dropped ball and run with it.
hammyhavoc··on Tim O'Reilly -Where Is AI on the Enshittification Curve?
Inappropriately choosing LLMs as a solution then forcing it onto employees, customers and anyone it directly or indirectly effects is not a curve—it is its own circle.

LLMs probably have some kind of value, but the immediate default of "can we solve this problem with an LLM?" from non-tech normies—and equal fetishization by techbros who present it as a false dichotomy of being pro or anti "AI"—highlights why managerial roles, and similar, shouldn't create tools that they themselves don't use. See developers making absolutely crap user interfaces for tools they themselves will never use and thus don't understand what makes it a good or bad solution because the tool isn't for their own job.

On that note, people saying the code output from LLMs is consistently high quality should ask their LLMs to define the Dunning-Kruger effect for them.

hammyhavoc··on I want an iPhone Mini-sized Android phone (2022)
With a battery that can be swapped rapidly without tools. Bonus points for pogo pins like a Samsung XCover phone.

Smaller size means smaller battery, but that's mitigated by the above. I want utilitarian. I don't want a phablet. I want practical and unobtrusive. The smartwatch was meant to replace the phone, but doesn't hit the right notes for me.

hammyhavoc··on Mark Cuban: Why AI Will Create More Jobs, Not Fewer
Ah yes, Mark Cuban—the same guy who said in 2017 that people shouldn't go and learn computer science as AI will be writing all the code within a few years and that people should instead go and get a liberal arts degree—on the basis that automating math is supposedly easier than automating critical thinking.

The same dude who said he would rather have bananas than Bitcoin then became another crypto chud in 2020 with crypto and NFT nonsense galore, because despite being a billionaire, fell for the FOMO.

The same dude who claimed that banks would collapse due to crypto and make it irrelevant—then the 2022 crypto Winter hit.

The same dude who also sold his Facebook shares shortly after the IPO.

The same dude who said that if Donald Trump won the 2016 election then the markets would crash.

Remember letting Steve Nash go in 2004? The same Steve Nash who then won two MVPs?

I think we can ignore Mark Cuban, and he should go and enjoy his billions. Quietly.

hammyhavoc··on In 2000 a Man Claimed to Be from 2036. Some Predictions Have Been Coming True
Monkeys and typewriters.

Some of what paranoid schizophrenics have come up to me and said in the street or on public transport has come true too—even what my own stepfather struggled with in terms of delusions—doesn't make it meaningful though. It's a statistical inevitability.

If a thousand people each pick a unique number between 2025 and 3025 for when the USA will enter into another significant major conflict or offer UBI then statistically some people will be correct but most people will inevitably be wrong.

Now extrapolate that to a larger sample size, include specific demographics of people who are extremely intelligent and previously had a firm grasp of reality before having a poorly managed psychiatric issue, and you will inevitably end up with some people (very tragically) being correct about what they're fearful or angry about.

hammyhavoc··on Thunderbird: Fluent Windows 11 Design
I want both via a single button-press, or defaults per-monitor.

What I want on an ultrawide isn't what I want on a portrait 16:9 side monitor.

hammyhavoc··on Thunderbird: Fluent Windows 11 Design
If you ever document it, I'd love to read about your workflow. me@hammyhavoc.com
hammyhavoc··on Thunderbird: Fluent Windows 11 Design
Something not making my eyeballs bleed is part-and-parcel to me actually wanting to use it. I value function over form, but Thunderbird has never been a looker. Plenty of UX friction too. It's just convoluted and messy.
hammyhavoc··on Thunderbird: Fluent Windows 11 Design
Is there any extension for Thunderbird that can handle "External Accounts" per-address signatures and sending identities already configured on a Gmail account?

I always go "I'll check out Thunderbird again" then "nope" out when I see it can't handle this kind of set up in the OOBE and most extensions don't receive ongoing support and thus stop being compatible.

I use Gmail on my phone and Pixel Watch, so ditching this setup is a non-starter as reconfiguring something as basic as email every time I get a new device or switch distros isn't my idea of a good time.

hammyhavoc··on LLM Inevitabilism
"Good technology sells itself."
hammyhavoc··on LLM Inevitabilism
What uses are you finding for it in the real world? I've found them nothing but unreliable at best, and quite dangerous at worst in terms of MCP and autonomous agents. Definitely not ready for production, IMO. I don't think they ever will be for most of what people are trying to use them for.

"Novelty" comes to mind.

← PreviousPage 3 of 34Next →