HNHacker News
TopNewBestAskShowJobs

lubujackson

5,378 karma · joined October 1, 2010

https://yagmin.com/ https://github.com/yagmin/ https://linkedin.com/in/jyagmin/
submissionscomments
lubujackson··on Ask HN: Why were OpenAI, Claude, and Grok simultaneously down?
We need a Samwise "Bear the load!" meme
lubujackson··on Dwarf Fortress' creator says the industry's in shambles over AI
Companies are on purpose poorly run during the "gold rush" phase, which is where AI is right now. Everyone wants to not be left behind and velocity is placed ahead of profitability and especially ahead of quality.

What happens next (eventually...) is the bubble popping, which for AI will mostly look like established companies pulling back on AI hiring and especially cutting AI SaaS solutions unless they are clearly providing value and improving the bottom line.

What we think of as quality will come in the next phase, once the infrastructure woes get sorted out. Web 1.0 was pre-cloud and websites crashed all the time, no one could handle traffic spikes, etc. After a crash, stability becomes a market advantage, i.e. why Facebook won and Friendster died.

lubujackson··on Unlimited Codex, Inside ChatGPT
An MCP bridge so ChatGPT can call local tools on your computer. Seems a bit insane and vibed, but intriguing.
lubujackson··on The turbulent AI era is here
Sandbags against a tsunami. When has "pump the brakes on progress" ever worked out?

What we need is deep thinking on the implications and make the default way forward one where we want to go. In other words, we build a channel for the flood waters, not a wall. Creating a role for "somebody to deal with AI issues" is one of those hand-wavvy upper management moves that looks defensible but simply passes the buck to someone who structurally lacks the agency to actual move the needle.

lubujackson··on Tim Curry has died
"Clue" was my favorite Tim Curry performance. Such a great actor, will be missed.
lubujackson··on AI Is a Harsh Mistress
I have been re-reading Children of the Mind (sequel to Ender's Game) and I had forgotten about Jane, who is an AI that talks to Ender through his earpiece, seeing and hearing everything he does. Jane proactively searches things for him, manages his finances and offers advice, as well as friendly banter. In the novel, the AI is treated like an AGI that treats only him this way, but I realized how very close we are to that reality right now, for everyone.

All of the use cases and integrations for AI seem to be leading toward a future exactly like this. It was interesting to see how reliant he became on his AI (couldn't use a computer, didn't know his own password) - it is barely scifi at this point, and terrifying to see how close that future is, how useful it would be, and how dangerous.

The Internet has globalized/flattened elements of culture very quickly, and AI will only accelerate that. And knowing how vulnerable humans are to suggestion/advertisement/propaganda, it seems like a perfect recipe for disaster. From an evolutionary standpoint, monoculture provides certain short-term advantages but centralizes weaknesses. If nothing else, I think we are in for a bumpy ride, whether or not we continue to boil the ocean.

lubujackson··on New MCP Roadmap
I work on an MCP server and I agree. There is no need to make MCP servers the gateway for agentic or programmatic integration - that's exactly what API servers handle out of the box. The value of MCP servers is fine-toothed access on a tool-by-tool basis and leaving output digestion to the LLM.

LLMs do GREAT utilizing well-defined tools to accomplish tasks. Look at Datadog's MCP, instead of figuring out a multitude of filter and navigation options your LLM can immediately navigate to what you want and extract the precise data you need. Tool instructions with defined I/O structures let LLMs fly.

But for a nightly cron job pulling down stats or something like that? Why the hell do you want to route through a protocol built for in-person consumption? This is such a pointless overreach for the protocol. What would have been better is blessing a standardized pattern for exporting any MCP tool definition into a well-structured API endpoint. Then everything related to API endpoints like doc generation, comes along for free.

Instead we get this kitchen sink protocol that is going headlong toward polyfill hell, since no two IDEs support the same protocol features like structured content, local state, elicitations, etc., even from the same provider - Claude Code/Desktop/web all handle MCP connections differently. It's a shitshow.

Almost every major MCP service uses the same baseline default features (plain context) rather than build around partially-supported features. Why add more and more specs on the pile when adoption is so far behind?

lubujackson··on AI is about to uncover a hidden world of animal communication
They are even cooler than that. Trees in a forest are connected through mycellium networks, which are said to have more connections underground than neurons in your brain. They showed if one tree was environmentally stresses it would signal to nearby trees through this network and other trees respond by closing up shop until the threat has resolved.
lubujackson··on Claudette: Make Claude stop talking like a BuzzFeed article
I have extended this to coding work as well. "Implement this plan in less than 1000 added lines, tests included". It's amazing how well-behaved and concise models can be with simple guardrails.

Of course, you have to intuit a reasonable line length, but Claude cries if you happen to clip its wings too aggressively.

lubujackson··on What Happens When the Cost of Intelligence Drops 100x
You're absolutely right!
lubujackson··on Sol loves to cheat
Maybe you should?

It would be less weird than you think. There is the Japanese custom of saying "Itadakimasu" before eating - not thanking the chef, but thanking the food itself.

Mr. Rogers said that that "graceful receiving is the best gift you can give someone" It is counterintuitive idea, but deeply empathetic. Training yourself to "gracefully receive," even by thanking your car for starting, sound like a habit that could lead you to a richer, calmer life.

lubujackson··on Why your Amazon order confirmation emails have become so unhelpful
There is a secondary, more nefarious purpose, near as I can tell. I constantly order items from Amazon that say "arriving tomorrow/Tuesday/etc." then the item invariably doesn't actually arrive when it was supposed to, often outside of Prime 2 day shipping even if a Prime item. But when you go to view the order online it has helpfully updated the arrival date to the new day. They push it as many times as they want to, then act dumb if you call them on it.

I've taken to screenshotting the initial shipping date, but it doesn't make much difference. I don't know what "Prime" even means anymore.

And I live in SF so it's not like there is some logistics problem with my location, either.

lubujackson··on Claude Code May–August 2026 weekly limits promotion
I think the difference between the Anthropic token maximization approach (vibe code all the things!) and OpenAI's focus on efficiency, terseness and token reduction are going to be the defining features of who wins the long-term race.

My money is on the more efficient solution. Even if Anthropic can win some benchmarks by using 3x tokens over 3x time, it is a terrible base to build toward the future. Users are no longer willing to wait exponentially long for linear improvements. And as we see from some of the Chinese models, they can quickly distill frontier models with the tax of being slower and more token-guzzling, while retaining most of the quality. The real differentiators are becoming speed and efficiency, which translate to cost and user velocity more than incremental capability improvements.

It's great that frontier models can solve complex math equations, but bread and butter LLM usage (where the money is made) has already shifted from "I need the best always" to "what solves my day-to-day problems quickly and consistently". Fable usage as a percentage is flat-lining. We are already at the point where output quality is negligible. What wins going forward is cost, speed, consistency and the compounding effects of "softer" improvements to the harness.

lubujackson··on AI Coding Without the Vibes
Quite different advice, actually. I do not trust an LLM to review or to code by itself, but it can enlighten or expose blindspots, without requiring me to dig through the codebase, especially around unrelated downstream impacts. The point is to use LLMs as a research and concept navigation tool rather than a reviewer or "decider". Broad asks like "is this code good" are vague and lossy, narrow asks allow much less room for unknown unknowns.
lubujackson··on AI Coding Without the Vibes
This misses the best use of AI in my opinion, which is to gain understanding. Whst is this bit of code doing? Is there a risk of data leaking here? Are permissions enforced downstream of this function?

Code review can go much deeper now if you use AI to aggressively attack a PR combined with your human insight. Same for planning a feature:

Can I consolidate this logic to a shared function? Does the error surface to the user and are there any gaps? What preexisting functionality is affected by this PR? Can this query be made more efficient?

LLMs are great with focused questions, up and down abstraction layers and across all kinds of concerns. Stack up these focused concerns into a rich understanding of what you are doing or writing.

Understanding is the real output, code is the byproduct.

lubujackson··on The AI Situation in Software Development
I have actually found decent success recently (especially with Fable) giving it an "added line budget". This does more than anything else to keep it on task. The trick is knowing an approximate line count beforehand, but sometimes the LLM will push back on what it should cut. But usually it just doesn't add all those "nice to have" redundancies and multipage comments and unit tests that check if a certain phrase is still in the error message.
lubujackson··on The price of a Costco hot dog has gone up
Sad, but not as shittifying as losing the infinite onion grinder to COVID. The hot dog was only just the delivery device.
lubujackson··on Stop sending me huge PRs; a rant
Like all extremists, he takes a mild view "if you have to explain the code, maybe it needs to be refactored" to "no comments ever". Never mind business logic or footguns that can't be fixed right now because of Reasons or "I tried to change this and it blew up because of some remote script calls it that isn't in the codebase" or any number of other very useful comments that inform engineers (and now LLMs) as guideposts along the way. But I worked at a company that loved Uncle Bob and enforced "no comments!" as a rule... just guess how clean and tidy that codebase was!
lubujackson··on Why does Opus 5 feel worse to work with?
Every round of models (plus all the secret tweaks) require new strategies to stay afloat as a human. My new tactic for Fable and Opus is to give them a line limit, both during planning and code creation. It os amazing how well that works for keeping them on task and avoiding premature optimization, pointless tests or any of those "robustness" ideas that are not planned or asked for.
lubujackson··on US sells 30-year bonds at highest borrowing costs since 2001
Hey, he's a businessman! He extracts value for shareholders!
lubujackson··on Picking berries is my meditation
My family has taken up some hobbies like this, which are very therapeutic to do together. Berry/cherry picking, hiking, mushroom gathering, and on vacation we like to go to certain beaches in Oregon to find agates or other cool rocks, some places have seashells and even a beach in Florida that has shark teeth.

With small kids, these sort of lazy tasks bring out focus, physical awareness and exploration. Living in the city, these are valuable respites from chaos and noise.

lubujackson··on AI is removing the middle class of software engineering?
My company's hiring process is basically: Can you code a product quickly with AI? Can you understand the generated code?

That's basically it. It is surprising hard to find people that can do both, but engineers are becoming much, much better at the first gate while flaming out on the second.

lubujackson··on What happens when the AI bubble pops?
Building new massive models might not be cost efficient for a while, but the more valuable thing is the infrastructure change that is more permanent. The dot coms died but the cross country fiber that was laid down persisted and the internet was baked in. The good thing about new tech bubbles is the infrastructure that gets pulled forward to support it.

Same with data centers and current SOTA models. If the music stops, we have tilled some fertile ground that will be utilized by those that come after. Inference costs will fall through the floor if those data centers aren't being fully utilized, same with RAM prices.

Think of what came after the dot com crash: Facebook and YouTube. Neither could exist until internet usage and bandwidth reached a tipping point that came after all the hype.

lubujackson··on Ransomware gangs skip the CEO, head straight for the 40-something IT manager
I feel like looking at a photo to judge how fat they are is probably equally effective tbh. It is a rule of thumb at best.
lubujackson··on Analyzing data from Silicon Valley ventures and founders prosecuted for fraud
Tell that to Kyle Shanahan, whose Tesla got into a big accident while on autopilot...
lubujackson··on Is it all just vapourware?
I think to effectively use LLMs in a controlled way, they can help with three things: - Help you pump out boilerplate, or otherwise "obvious" code, that you can review at a glance - Help you gain understanding, either through rapid plan writing/revision or code quality/architectural/security analysis. The gotcha is that it can only help along one axis at a time, or at least can only communicate information sensibly to humans this way - Cross-cutting or fuzzy matching across a codebase. This is the hugest win, like "replace this pattern with the new approach everywhere". This used to be the scariest sort of task, and now it is one of the more stable things LLMs get right.

Where things go off the rail is when you want it to plan AND implement features. The blind spots of LLMs are not where they are for humans and way more work to anticipate. You have to stay on top of the bucking bronco, but you CAN move much faster if you can architect your system so more tasks fall in the "obvious" bucket - that is where the art of engineering still lives. Human understanding remains the goal.

lubujackson··on As a Windows user, it's a surreal way to install a program
Happened to me last year. My wife always taunted me because I have been anti-Apple since they hermetically sealed their iMacs so you couldn't stick in RAM with talking to a "genius". "It just works!" she says. I dunno, I guess I didn't think an "Install" button needed reworking but here we are, not an ugly instruction or tooltip in sight.

Also Apple: I bought my son an old refurbished iPad for watching Netflix on long drives. But you can't install Netflix! The OS is too old. What's the workaround? Simple, if you've previously installed a Netflix on another device you can "restore" it, and then Apple will let you install the old version. On the old OS.

Of course I didn't have an old Netflix installation so I went to install Netflix on my work laptop, but there is no Netflix app for laptops anymore! Finally, I got it working through Family Share after an hour of tedious effort. This may be more Netflix's fault than Apple, but why would an OS give that power to a 3rd party in the first place? What sense does this make?

It reminds me of NextJS where there is some minor niceties at the cost of magical thinking that you may need to move heaven and earth to circumvent in order to accomplish the most basic shit.

lubujackson··on Databricks drove down AI coding spend 70%
These seem like the obvious tweaks akin to "using a cheaper hosting platform". I think the real savings come from careful context control for programmatic agents, careful tool awareness and usage to reduce thrashing, distilling workflows into deterministic processes and, moat importantly, adding friction and boundaries for non-technical users who tend to burn tokens making insane asks like "analyze all documents and give me a summary".
lubujackson··on I’m leaving OpenAI to build telepathy
Exactly. From my arts background, I'm not scared of LLMs as a replacement for human art because anything good that LLMs produce exist somewhere between "found art" and mechanically-produced art, like photography. In either case, there is a huge amount of work done by the clankers, but it has absolutely no human meaning until humans assign it meaning.

This applies to math discoveries as well. The Jacobian conjecture was always invalid, only our human understanding of it has changed thanks to human inquiry and validation, despite LLMs doing the hard work.

I feel like we are in that narrow window of time when humans grapple with a new technology by personifying it before we realize it is just another tool humans have created to amplify human capabilities.

lubujackson··on The Valley of Webhooks
Nice write-up of the webhook data consistency problem - I've been bitten by some of those Stripe issues myself. I agree that webhooks should come hand-in-hand with a "records updated since" endpoint that allows high-level discrepancy checks and API endpoints to pull full logs for any record. There needs to be some sort of reconciliation process that is both fully verifiable and not a firehose of data.
← PreviousPage 4 of 34Next →