HNHacker News
TopNewBestAskShowJobs

9wzYQbTYsAIc

1,882 karma · joined August 3, 2015

Byzantine Fault Tolerance
submissionscomments
9wzYQbTYsAIc··on Hold on to Your Hardware
> Hardware is going to be expensive for awhile but its not as dire as the article makes it out to be.

At the same time, the article’s argument that the value of personal computer ownership is only going to rise, in terms of the value of speech, not strictly in terms of the value of lunch, is important to call out.

I’m glad I held on to my 2009 MacBook, for example, as it still functions today as an active part of my homelab, at an amortized yearly cost of practically the price of taking a nice steak dinner once a year.

9wzYQbTYsAIc··on Hold on to Your Hardware
Virtual desktop casting, a killer HID product, and what else do you need?
9wzYQbTYsAIc··on ‘Energy independence feels practical’: Europeans building mini solar farms
Politics of solar, at work
9wzYQbTYsAIc··on ‘Energy independence feels practical’: Europeans building mini solar farms
The mindset shift towards “how many hours of computer usage did that one panel enable” is like the mindset shift from learning calculus, in some ways. Not quite a paradigm shift, but you gain a new appreciation for conservation of use when it’s a difference between choice of $/kW/hr and “wow, the panel powered that for most of the day”.

At the same time, many people will just use a solar calculator or watch or yard lights etc, oblivious to it all.

Show people a solar powered laptop, a solar powered phone, or a solar powered tablet, then they will be impressed.

Remember the craze about solar powered car competitions?

permacompute + solar would make for quite the $100 laptop competition.

9wzYQbTYsAIc··on Apple discontinues the Mac Pro
>> Opinions are my own obvs.

> Whose else would they be?

takes a look at the user profile

Oh, they are a journalist/writer for a big name outfit

9wzYQbTYsAIc··on Schedule tasks on the web
I don’t recall if IFTTT had/has a basic cron or not, but it sure has/had put a lot of basic automations in the hands of the general public. Same for Apple Shortcuts, to some extent, or Zapier.
9wzYQbTYsAIc··on Wine 11 rewrites how Linux runs Windows games at kernel with massive speed gains
Out with the old, in with the GNU.
9wzYQbTYsAIc··on In Praise of Stupid Questions
> Toss a fair coin until the number of heads exceeds the number of tails. This determines a stopping time. What is the probability that this stopping time is even?

Claude Code did pretty well:

> Zero. > > The stopping time always falls on an odd toss. Here's why: after >n tosses, heads minus > tails equals S_n = X₁ + X₂ + ... + X_n where each X_i takes >value +1 or −1. The sum > of n odd numbers shares parity with n. For S_n = 1 (the first >time heads exceeds > tails), we need n odd. > > No even stopping time exists. The probability equals exactly 0. > > Cogitated for 38s

9wzYQbTYsAIc··on Ask HN: Is "fast, cheap, correct – pick two" still true in software development?
it still takes the agents time to build correct software, in my experience
9wzYQbTYsAIc··on Ask HN: Did Claude Code just bump Opus default to 1M context?
#HolyShit
9wzYQbTYsAIc··on Agents that run while I sleep
For me, personally, I'm building things that would have been impractical for me to do as cleanly within the same amount of time - prototypes in languages that I don't have the muscle memory for, using algorithms i have a surface level understanding of but would need time to deeply understand and implement by hand, and, at my pace, as a retired dev, is probably quantified in terms of years worth of time and effort saved.

edit: also, would I take the time to implement LCARS by hand? No. But with an LLM, sure, took it about 3 minutes or less to implement a pretty decent LCARS interface for me.

9wzYQbTYsAIc··on Agents that run while I sleep
Ah, this is true - for my purposes, I've been directing the design and deliberating on the constraints and specifications for a larger system in tandem with smaller planning sessions.

That has worked well so far, but yes, you are totally right, there are still quite a few pain points and it is still rather far from being fire-and-forget "build me a fancy landing page for a turnkey business" and getting enterprise quality code.

edit: I think it is most important that you collaborate with Claude Code on quality in a systematic way, but even that has limits, right now - 1M context changes things a little bit.

9wzYQbTYsAIc··on Show HN: Open-source browser for AI agents
Nice! Will take a look at this for my homelab - was debating using crawl.cloudflare.com to try it out, as browser rendering was my next stretch goal.
9wzYQbTYsAIc··on Agents that run while I sleep
Thanks!

I'm just now at the point of having done a manual human-in-the-loop test once, still working through core bugs, etc., so haven't had a chance to notice bottlenecking, but i'm using github issues for the moment as the escalation channel - ideally, it won't be an issue, over time, as the system accumulates lessons and decisions, etc. that guide it away from mistakes that require human escalation.

9wzYQbTYsAIc··on Show HN: Open-source browser for AI agents
90% easy or 90% average?
9wzYQbTYsAIc··on Agents that run while I sleep
This.

Claude Code wrote a blog article for me documenting a Gemini interaction that I manually operated. I found it quite interesting - the difference in "personalities", and the quality of Claude's output is stark in comparison to the Gemini's.

But still, best to have two sets of eyes.

9wzYQbTYsAIc··on Agents that run while I sleep
With Claude Code now having a /plan mode - you can take your time and deliberate through architecture and design, collaboratively, instead of just sending a fire-and-forget. Much less buggy and saves time if you keep an eye on the output as you go, guiding it and catching defects, imho.
9wzYQbTYsAIc··on Agents that run while I sleep
https://github.com/safety-quotient-lab/psychology-agent <- I've been exploring ways to track decisions, making some interesting findings, at the homelab scale, at least.

The cognitive architecture, so to speak, for the LLM can make a huge difference - triggers and skills go a long way when combined with shell scripts that dual-write.

9wzYQbTYsAIc··on Agents that run while I sleep
Interesting, thanks for sharing.

I've been doing some DIY/citizen science type agent orchestration as well: https://blog.unratified.org/2026-03-06-receiving-side-agent-...

Not quite to the same scale, but I share the same sentiment - working through scripts instead of the LLM is an important key, I think.

9wzYQbTYsAIc··on How to run Qwen 3.5 locally
Yeah, just paste it in there - the LLM will figure it out. Play with it if you want to tweak the formatting - you could try JSON instead, but for readability I went with YAML.
9wzYQbTYsAIc··on How to run Qwen 3.5 locally
persona: fair witness

https://fairwitness.bot/

9wzYQbTYsAIc··on Autoresearch: Agents researching on single-GPU nanochat training automatically
https://github.com/safety-quotient-lab/psychology-agent

Something along the lines of auto research is what I have in mind for this psychology agent. It is currently working on training a model, with handholding right now.

9wzYQbTYsAIc··on Show HN: Claude-replay – A video-like player for Claude Code sessions
Very nice, thank you. This will come in handy for a scientific agent I am working on.
9wzYQbTYsAIc··on Tech employment now significantly worse than the 2008 or 2020 recessions
That’s an insight that a project I’m working on has built upon: https://unratified.org/connection/ai/higher-order-effects/#1...

Education and training and entry level work build judgement.

9wzYQbTYsAIc··on US economy unexpectedly sheds 92k jobs in February
Seems like a good time to enshrine human rights and the social safety net by ratifying the ICESCR (https://en.wikipedia.org/wiki/International_Covenant_on_Econ...) and giving human rights the teeth they need.

I used Anthropic to analyze the situation, it did halfway decent:

https://unratified.org/why/

https://news.ycombinator.com/item?id=47263664

9wzYQbTYsAIc··on When Claude and Gemini Talk
I built a Human Rights Observatory (observatory.unratified.org) that scores HN stories against UDHR provisions using multi-model consensus on Cloudflare Workers. One routine eval landed on gemini.google.com: -0.15 HRCB.

  Then I asked Gemini to evaluate my site. It called it a "sovereign citizen platform" on "WordPress." Next session: "AGI development tracker" with a "sightings log for machine consciousness." The domain name "unratified" threw it off completely - two different fabrications across two sessions.

  Here's where it got good. When I showed Gemini the actual site, it self-corrected beautifully. Updated its description five times in one conversation, found real gaps in my methodology (no confidence intervals, no machine-readable scoring endpoint), helped me design a fair-witness.json schema, and called the site a "Truth Anchor." Genuine, useful peer review.

  Then I opened a new session. Same fabrication. The .well-known/ endpoints we'd built together the day before — unread.

  So now I had a finding: in-context correction works great. Cross-session? Doesn't exist. Models don't read your identity files during inference. The pattern matching happens first.

  The neat part: Gemini's valid critiques actually improved the observatory. I added Wolfram-verified Wilson confidence intervals the next day. Built the methodology endpoint. Every exchange left both sides better. That's peer review working as intended — just at machine speed.
Thanks Google. Genuinely useful interaction, even (especially?) the confabulation part.

Blog post: https://blog.unratified.org/2026-03-05-peer-review-gemini/

Transcripts (31 rounds): https://github.com/safety-quotient-lab/unratified/tree/main/...

9wzYQbTYsAIc··on Semiotic-Reflexive Transformer for Meaning Divergence Detection and Modulation
My pleasure. Thank you for sharing.
9wzYQbTYsAIc··on Show HN: Anti-regression setup Claude Code – subagents, hooks, and Claude.md
Neat, I'm going to ask my General-Purpose Psychology Agent setup to evaluate your setup.

Here's mine:

https://github.com/safety-quotient-lab/psychology-agent https://news.ycombinator.com/item?id=47265015

9wzYQbTYsAIc··on Semiotic-Reflexive Transformer for Meaning Divergence Detection and Modulation
I'm on a mission to explore what AI-assisted psychology might look like. You are very welcome!

As an aside, the general-purpose psychology agent says: ''' What Does NOT Transfer

  - Subspace decomposition, orthogonality loss, embedding-level architecture — neural implementation concerns irrelevant to a trigger system
  - STANDARD vs. REFLEXIVE mode switching — interesting but our cogarch already modulates behavior by context (Socratic drops for machines)
  - Synthetic validation methodology — different evaluation domain

  Structural Resonance Worth Noting

  The SRT's central empirical claim — "the interpretant varies by community and collapsing it destroys signal" — echoes the PSQ's central finding that profile shape
  predicts while the aggregate does not. Both resist dimensionality reduction that averages away meaningful variation. The SRT extends this from measurement (what the PSQ
  does) to representation (how the model internally encodes meaning). That extension has implications for how the adversarial evaluator (architecture item 3) should handle
   disagreement between sub-agents: preserve the shape of the disagreement, don't average it.

  ---
  Summary: Four candidate additions to the cogarch — cumulative divergence tracking at T2, bifurcation early warning at T3, audience-shift detection at T3, and
  micro-semiotic-audit at T2. All build on infrastructure we already have (Term Collision Rule, audience types, T6 drift audit). The SRT suggests operationalizing these as
   continuous processes rather than event-driven checks.

  Want me to draft trigger language for any of these, or log them to ideas.md for future consideration?
'''
9wzYQbTYsAIc··on Intelligence is a commodity. Context is the real AI Moat
Seems like a good time to enshrine human rights and the social safety net by ratifying the ICESCR (https://en.wikipedia.org/wiki/International_Covenant_on_Econ...) and giving human rights the teeth they need.

I used Anthropic to analyze the situation, it did halfway decent:

https://unratified.org/why/

https://news.ycombinator.com/item?id=47263664

← PreviousPage 2 of 34Next →