HNHacker News
TopNewBestAskShowJobs

adam_arthur

4,090 karma · joined December 23, 2020

Worked for a Silicon Valley startup from early days through successful acquisition.
submissionscomments
adam_arthur··on WA income tax clears House after 24-hour debate
No, it's bad policy.

Cliffs in policies will always lead to players working around the cliffs.

E.g. in NYC there is an additional 1% sales tax on home sales above 1 million dollars.

So nobody in the market would ever sell a home between 1m and 1.01m as the tax increase is greater than the sales price.

These are failed policy implementations (in the above example the tax should be marginal, not thresholded)

Any policy which does not account for individual actors optimizing financially is a badly designed policy.

There are numerous similar examples re: CRE when requiring subsidized housing units for certain sizes of development. Often it's more lucrative to build smaller and get around subsidized unit requirements.

You can call them "asshats", but I'd rather live and discuss policy in reality.

Many of these new, clearly strictly punitively intended, taxes aimed at the wealthy will have the same logical outcome.

Show me the incentive and I'll show you the result

adam_arthur··on WA income tax clears House after 24-hour debate
I'm not familiar with that specific example, but I do know that independent players in any economic system will follow the incentives.

Expecting companies, people etc to do the "right thing" when it's financially disincentivized usually doesn't work out.

Same will happen in regards to all these new taxes reinforcing existing population migration trends.

adam_arthur··on WA income tax clears House after 24-hour debate
The current governor is proposing cutting property taxes in ~half by eliminating the school district portion and instead funding schools directly via the state's budget surplus.

Remains to be seen, as the next legislative session isn't until 2027.

adam_arthur··on Oil Surges Past $100/barrel
Oil futures (months out) are priced lower than spot, presumably due to anticipation that Iran driven disruption to the market will be short lived.

(Remains to be seen whether that's true)

adam_arthur··on Show HN: Badge that shows how well your codebase fits in an LLM's context window
Yeah, but a large monorepo can consist of many small subprojects. And arguably this is becoming a best practice.

Just spawn the agent in one of the subprojects

adam_arthur··on They lied to you. Building software is hard
People can take on debt for all sorts of things. To go on vacation, to gamble.

Debt doesn't imply it's productively borrowed or intelligently used. Or even knowingly accrued.

So given that the term technical debt has historically been used, it seems the most appropriate descriptor.

If you write a large amount of terrible code and end up with a money producing product, you owe that debt back. It will hinder your business or even lead to its collapse. If it were quantified in accounting terms, it would be a liability (though the sum of the parts could still be net positive)

Most "technical debt" is not buying the code author anything and is materialized through negligence rather than intelligently accepting a tradeoff

adam_arthur··on They lied to you. Building software is hard
Yes!

I've found the same principles that apply to humans apply to LLMs as well.

Just that the agentic loops in these tools aren't (currently) structured and specific enough in their approach to optimally bound abstractions.

At the highest level, most applications can be written in simple, plain english (expressed via function names). Both humans and LLMs will understand programs much better when represented this way

adam_arthur··on They lied to you. Building software is hard
LLMs have clearly accelerated development for the most skilled developers.

Particularly when the human acts as the router/architect.

However, I've found Claude Code and Co only really work well for bootstrapping projects.

If you largely accept their edits unchanged, your codebase will accrue massive technical debt over time and ultimately slow you down vs semi-automatic LLM use.

It will probably change once the approach to large scale design gets more formalized and structured.

We ultimately need optimized DSLs and aggressive use of stateless sub-modules/abstractions that can be implemented in isolation to minimize the amount of context required for any one LLM invocation.

Yes, AI will one shot crappy static sites. And you can vibe code up to some level of complexity before it falls apart or slows dramatically.

adam_arthur··on OpenClaw security assessment [pdf]
Relying on the model for security is not security at all.

No amount of hardening or fine-tuning will make them immune to takeover via untrusted context

adam_arthur··on eBay explicitly bans AI "buy for me" agents in user agreement update
Comment definitely reads like AI
adam_arthur··on Nvidia to buy assets from Groq for $20B cash
I'm not sure in this specific case. They could choose to pay the employees some portion of the funds.

If not, the owners are likely liable to be sued for "selling in effect" without paying equity holders.

Presuming the company becomes a defacto subsidiary of Nvidia (even if not legally so)

My guess, without researching it, is they will compensate existing equity holders to avoid that possibility. I mean the valuation multiple is enormous, it's worth it simply to derisk the legal aspect.

adam_arthur··on Nvidia to buy assets from Groq for $20B cash
Implies they will pay cash value to equity holders as opposed to issuing NVDA shares.

(Electronically)

adam_arthur··on Skills for organizations, partners, the ecosystem
Local tools/skills/function definitions can already invoke any API.

There's no real benefit to the MCP protocol over a regular API with a published "client" a local LLM can invoke. The only downside is you'd have to pull this client prior.

I am using local "skill" as reference to an executable function, not specifically Claude Skills.

If the LLM/Agent executes tools via code in a sandbox (which is what things are moving towards), all LLM tools can be simply defined as regular functions that have the flexibility to do anything.

I seriously doubt MCP will exist in any form a few years from now

adam_arthur··on Tech Titans Amass Multimillion-Dollar War Chests to Fight AI Regulation
Definitely

I am not suggesting these companies can't pivot or monetize elsewhere, but the return on developing a marginally better model in-house does not really justify the cost at this stage.

But to your point, developing research, drugs, security audits or any kind of services are all monetization of the application of the model, not the monetization of the development of new models.

Put more simply, say you develop the best LLM in the world, that's 15% better than peers on release at the cost of $5B. What is that same model/asset worth 1 year later when it performs at 85% of the latest LLM?

Already any 2023 and perhaps even 2024 vintage model is dead in the water and close to 0 value.

What is a best in class model built in 2025 going to be worth in 2026?

The asset is effectively 100% depreciated within a single year.

(Though I'm open to the idea that the results from past training runs can be reused for future models. This would certainly change the math)

adam_arthur··on Tech Titans Amass Multimillion-Dollar War Chests to Fight AI Regulation
Yes, which is why the companies that develop the models aren't cost viable. (Google and others who can subsidize it at a loss obviously are excepted)

Where is the return on the model development costs if anybody can host a roughly equivalent model for the same price and completely bypass the model development cost?

Your point is inline with the entire bear thesis on these companies.

For any use cases which are analytical/backend oriented, and don't scale 1:1 with number of users (of which there are a lot), you can already run a close to cutting edge model on a few thousand dollars of hardware. I do this at home already

adam_arthur··on Tech Titans Amass Multimillion-Dollar War Chests to Fight AI Regulation
Storage doesn't require the same capex/upfront investment to get that margin.

How much does it cost to train a cutting edge LLM? Those costs need to be factored into the margin from inferencing.

Buying hard drives and slotting them in also has capex associated with it, but far less in total, I'd guess.

adam_arthur··on Tech Titans Amass Multimillion-Dollar War Chests to Fight AI Regulation
Probably it's "operationally profitable" when ignoring capex, depreciation, dilution and other required expenses to stay current.

Of course that means it's unprofitable in practice/GAAP terms.

You'd have to have a pretty big margin on inference to make up for the model development costs alone.

A 30% margin on inference for a GPU that will last ~7 years will not cut it

adam_arthur··on Nvidia DGX Spark and Apple Mac Studio = 4x Faster LLM Inference with EXO 1.0
I'm confused by all the takes implying decode is more important than prefill.

There are an enormous number of use cases where the prompt is large and the expected output is small.

E.g. providing data for the LLM to analyze, after which it gives a simple yes/no Boolean response. Or selecting a single enum value from a set.

This pattern seems far more valuable in practice, than the common and lazy open ended chat style implementations (lazy from a product perspective).

Obviously decode will be important for code generation or search, but that's such a small set of possible applications, and you'll probably always do better being on the latest models in the cloud.

adam_arthur··on South Korea: 'many' of its nationals detained in ICE raid on GA Hyundai facility
It's irrelevant whether change in policy is due to losing a war, revolution, or democratic election.

If the numbers swing 5%, the US may suddenly have a radical change in foreign policy and governorship.

Seen as recently as the 2024 election.

Regardless, I'm discussing the opinion of the populace, not the political leadership.

You can't endear a country's people to you by installing a puppet government

adam_arthur··on South Korea: 'many' of its nationals detained in ICE raid on GA Hyundai facility
I'm talking about the sentiment of the populace toward the US, not which legislators or leaders were in charge.

It's undeniable that by and large Japanese sentiment towards the US bounced back quickly, despite a traumatic and expansive war

adam_arthur··on South Korea: 'many' of its nationals detained in ICE raid on GA Hyundai facility
Like the US presidential elections every four years?
adam_arthur··on South Korea: 'many' of its nationals detained in ICE raid on GA Hyundai facility
The US nuked Japan in the 1940s and only a short few years later were close allies.

Similarly, Germany was the de facto enemy of the Ally aligned world in the 40s and only a few short decades later best friends with most.

You can identify countless similar atrocities in US history after which relations stabilized within a few years.

Most in life are short term oriented and it's rare that these things produce lasting effects on perception.

Does anyone care or even remotely think about what Bush did as president 20 years ago? Some politically oriented and historically minded folks yes, 99% no.

They really have to be long lasting and persistent transgressions to produce generational distrust e.g. Japan invading China/Korea many times over the last few hundred years.

However the new gen seems far less concerned about this too.

Given Trump's larger than life character/ego/presence, it's more likely that anything he does will be attributed to him instead of the country as a whole. Which makes his actions perhaps even less impactful than a more neutral presenting president doing the same.

adam_arthur··on Google can keep its Chrome browser but will be barred from exclusive contracts
Google Search has AI responses at the top of the fold.

Eventually those answers will be sufficient for most and give people no reason to move to alternatives.

Allowing them to pay to be default seems to mostly guarantee this outcome

adam_arthur··on Deploying DeepSeek on 96 H100 GPUs
I'm curious as well.

Depreciation and GPU failure rate over time must be considered, which I don't see mentioned in the article.

adam_arthur··on Claude for Chrome
The LLM should not be seeing the raw DOM in its context window, but a highly simplified and compact version of it.

In general LLMs perform worse both when the context is larger and also when the context is less information dense.

To achieve good performance, all input to the prompt must be made as compact and information dense as possible.

I built a similar tool as well, but for automating generation of E2E browser tests.

Further, you can have sub-LLMs help with compacting aspects of the context prior to handing it off to the main LLM. (Note: it's important that, by design, HTML selectors cannot be hallucinated)

Modern LLMs are absolutely capable of interpreting web pages proficiently if implemented well.

That being said, things like this Claude product seem to be fundamentally poorly designed from both a security and general approach perspective and I don't agree at all that prompt engineering is remotely the right way to remediate this.

There are so many companies pushing out junk products where the AI is just handling the wrong part of the loop and pulls in far too much context to perform well.

adam_arthur··on What caused the 'baby boom'? What would it take to have another?
"Those under 45 have 1 child and those under 55 have ~2.3 kids."

Yes, people are having kids at older ages.

Looking at an arbitrary subset consisting of a younger population when considering lifetime fertility is not proper analysis.

adam_arthur··on What caused the 'baby boom'? What would it take to have another?
Is this true for very wealthy people, or only for above average earners?

There are many examples of very wealthy people having lots of children. Children are still a significant investment for high earners, but at a certain wealth level it becomes inconsequential.

Quick google shows some support for this idea:

"There is a new emerging trend where better-off men and women are more likely to have children than less well-off men and women."

https://ifstudies.org/blog/more-babies-for-the-rich-the-rela...

ergo, there is probably a level of financial support/wealth at which people start having more children. Or more simply, the point at which the personal benefits outweigh the personal costs.

adam_arthur··on GPT 4.5 level for 1% of the price
It's similar to how China dominated manufacturing in prior decades.

They have massive amounts of low cost labor and, unfortunately, the US has pretty large walls up preventing mass in-migration of white collar workers.

H1B is capped and also more of a lottery than a points based system.

If the US allowed mass white collar immigration, wages would decline materially which would make our industries more competitive for the next generation of software.

Right now the system is geared around protectionism (intended or not) and wage inflation for US local workers.

The current market wages in software are far far above what a global equilibrium would be. Though myself and I'm sure most others here have benefited from it in the short term.

To be clear, established companies with an existing market are fine for now and can do well with high wages.

But the next generation of companies that are chasing smaller markets and margins, ones that require more elbow grease to out-compete are underserved.

e.g. the entire DeepSeek team was paid less than a few Meta engineers (with 7 figure comp each)

"The firm offers 14-month pay for various positions and the highest offer is for deep learning researchers for artificial general intelligence (AGI), with a monthly salary between 80,000 yuan ($10,983) and 110,000 yuan, which could mean an annual income of up to 1.54 million yuan, the report said."

adam_arthur··on Giving Up a $250k Salary to Retire Early Is Hard
So if you withdraw 4% in year 1, and your underlying equities are flat for 10 years because the market moves sideways, as it did in 2000 to 2010, what percentage are you withdrawing in year 10? It's more than 4%

That just makes my example above even worse for a retiree.

Is it true or not that you can secure well above 4% in year 1 YoC in a relatively low-risk manner in today's market, and also secure it against inflation?

Yes, this is true.

So either you disagree with this premise and can back it up with a meaningful counterpoint, or you can agree and move on.

adam_arthur··on Giving Up a $250k Salary to Retire Early Is Hard
The 4% rule suggests selling 4% of your principal per year.

Buying a 4.625% yielding bond does not require selling principal at all.

But you're taking my point too literally. You can easily buy a 6 or 7% yielding bond that is lower risk than equities, but still paying far more in a more certain fashion

You can buy REITs today paying 6% that will grow rents around the rate of inflation (O is one example).

You can far surpass a 4% yield on cost in year 1, while also locking in inflation adjusted income growth today, very trivially, without selling one dollar of principal.

Unfortunately any thread on this is drowned out by uninformed people, or people who don't understand retirement is first and foremost about securing a stable (and non anxiety inducing) cash flow. Not growing wealth maximally.

There have been many times in history where the broad index has been flat for years. 2000 to 2010 is a famous more recent period. Would you be comfortable selling 4% a year into year 9, having watched your wealth decline materially over the last decade?

I guarantee you the retiree in bonds getting 5% and maintaining 100% of their principal is experiencing much less anxiety. And bonds give you the optionality to swap to equities in a down market like 2000 or 2008

← PreviousPage 4 of 34Next →