Google 2025 recap: Research breakthroughs of the year

146 points by Anon84 10 hours ago on hackernews | 110 comments

7373737373 | 8 hours ago

I wonder if/when DeepMind will try to tackle the problem of finding potential room temperature, ambient pressure superconductors (and if it can be approached similarly to protein folding)

jiggawatts | 8 hours ago

Not enough training data and not enough theory for synthetic data, sadly.

zkmon | 8 hours ago

They should call it as very specific to AI, instead of general research. How can it be an "Year of agents", when agents haven't stepped out of the programming work?

pfortuny | 4 hours ago

When you only have a hammer...

fancyfredbot | 7 hours ago

Google are really firing on all cylinders recently. It's almost shocking to read all they've done in the last year.

The fact they caught up with OpenAI you almost expect. But the Nobel winning contributions to quantum computing, the advances in healthcare and medicine, the cutting edge AI hardware, and the best in class weather models go way beyond what you might have expected. Google could have been an advertising company with a search engine. I'm glad they aren't.

10xDev | 7 hours ago

Meanwhile the economy is tanking. But yeah what a fantastic year it is to be a company worth trillions.

cpursley | 7 hours ago

You don’t believe the recent economic numbers? I’m not disagreeing with you, just curious about other takes (and generally very skeptical of funny money printer go burrr economic things vs real economy meaning real output).

MrOrelliOReilly | 6 hours ago

There has been a lot of discussion on this recently in the blog-o-sphere. All conclusions I've seen so far are that the economy is basically fine and maybe people's expectations have risen (I'm oversimplifying). I'm also quite eager to hear different conclusions, because there is a lot of cognitive dissonance on the economy right now.

- https://www.slowboring.com/p/you-can-afford-a-tradlife

- https://www.slowboring.com/p/affordability-is-just-high-nomi...

- https://thezvi.substack.com/p/the-revolution-of-rising-expec...

- https://open.substack.com/pub/astralcodexten/p/vibecession-m...

hansvm | 3 hours ago

There's definitely an aspect of rising expectations (e.g., everyone and their dog having a late-model smartphone). There's also an aspect where some of that is mostly unavoidable (e.g., accessing my HSA now requires a very late-model smartphone -- something I can avoid complying with for now by just finding a better place to transfer my money, but it's a worrying trend -- to achieve the same QoL as in the early 2000s I have mandatory nontrivial overhead).

It's really not just that though. A lot goes into it, but one observation is that the relative increases in wages and prices isn't distributed evenly. Some examples:

- A lot of people are legitimately substantially better off than they would have been a few decades ago. I literally never have to worry about money anymore when thinking about our purchases (for everything but a house with a big yard, which we still can't safely [0] afford without moving). I'm not alone.

- That's not true of everyone, even my next-door neighbors. I know people splitting a studio apartment and still struggling a bit. They have good jobs, and even splitting the apartment their post-tax, post-rent pay is $7.20/hr. That's fine enough I suppose, but they'll literally never be able to save for a home of any quality in the area in their entire lives using only a single income. It'll take them awhile to afford a home anywhere.

- Suppose you have a couple young kids. That places hard bounds on how much money you need to make even for childcare to make sense to get up to two incomes in the first place. I've known plenty of people with PhDs and good jobs who quit to take care of the kids for financial reasons, supporting the household on just the higher-earner's pay.

- A lot of small towns haven't seen the same increase in wages as the rest of the country but have seen the increase in prices. My hometown saw an increase from $10/hr to $20/hr in what a great wage is over the last 25 years. CPI only went up 1.9x in that time, but the same caliber of house went up 3x, and the staples people used to eat (like ground beef) went up more than 3x as well. They're correctly observing that they have less take-home money (because of 3x increased rent), that take-home money doesn't go as far (they can't eat the same foods they could 25yrs ago), and it definitely doesn't go as far if you want to do something like save for a house (it's an extra 4+yrs of post-tax, post-rent income to pay for a house, assuming you could devote all of it to savings instead of groceries and whatnot).

I'm not sure exactly how to quantify who's struggling and why at a macroscopic level, but I guarantee they're real and that it's not just an increase in expectations.

[0] It depends on your relative risk levels, but if you're not convinced the gravy train will last forever and are concerned about locking up all your assets in a depreciating vehicle then you need to be a bit more frugle with your choice of home.

10xDev | 6 hours ago

Truth be told, it is more complicated than a "tanking" economy so you will get headlines like this: https://www.bbc.com/news/articles/c62n9ynzrdpo but that's because it is a K-shaped economy: https://www.theguardian.com/business/2025/dec/07/stock-price... thanks to the stock market and AI investment. The job market especially for entry level tech jobs is also essentially screwed whether that's due to AI or something else, who even knows anymore: https://www.aljazeera.com/economy/2025/12/16/us-unemployment...

inerte | 5 hours ago

Vibecession so good I remember we’ve been a quarter away from recession for the last decade.

fancyfredbot | 6 hours ago

I'm not sure that the rest of the economy really is "tanking" but OK. Are you implying it's distasteful to discuss success from a big company in such dark times?

Google could really easily be a purely rent seeking business but they are innovating, and if you are worried about the economy then this should seem like good news.

andsoitis | 6 hours ago

> Meanwhile the economy is tanking.

NYT: US GDP Grew 4.3%, surging in 3rd Quarter 2025 - https://www.nytimes.com/2025/12/23/business/us-economy-consu...

WSJ: Consumers Power Strongest US Economic Growth in 2 years - https://www.wsj.com/economy/us-gdp-q3-2025-2026-6cbd079e

The Guardian: US economy grew strongly in third quarter - https://www.theguardian.com/business/2025/dec/23/us-economy-...

Meanwhile, consumer debt is at record highs.

https://www.newyorkfed.org/microeconomics/hhdc

andsoitis | 5 hours ago

> consumer debt is at record highs.

While consumer debt is at or near historical highs, it is in and of itself not a problem (broader economic risk).

What you need to look at as well is debt burden ratios and repayment behavior, not just raw totals.

Household debt service ratio (the share of disposable income spent on principal + interest payments) is well below historical crisis peaks (e.g., 2007–2008), suggesting households are currently spending a smaller share of income on debt payments than in past stress periods.

While total household debt is at record levels (~$18 trillion+), debt as a share of income or GDP has not reached past crisis peaks like 2008. That means debt growth hasn’t outpaced income growth as dramatically as in previous crises.

However, delinquency rates, especially for credit cards and student loans, are elevated, nearing or exceeding long-run highs outside recessions.

Mortgage delinquency rates remain lower than unsecured debt categories, but have ticked up slightly. Because they're relatively stable, it mutes broader systemic risk for now.

jeffbee | 5 hours ago

And you didn't even mention the population.

andsoitis | 4 hours ago

> And you didn't even mention the population.

household debt per capita is also trending up, so larger population is not the driver of increased consumer debt.

jeffbee | 4 hours ago

It is nonetheless true that to interpret such a chart as the one the GP posted you must at least mentally discount it for both population (which is +11% since 2008, the last consumer credit calamity) and the value of dollars (which are now ~67¢ vs. 2008). Debt service as fraction of HH income is in some ways easier to interpret.

Anyway, even clicking through to the PDF linked from GP's front page shows that every metric of US consumer credit is at or near all-time bests.

esseph | 4 hours ago

Car loans.

"The percentage of subprime borrowers – those with credit scores below 670 – who are at least 60 days late on their car loans has doubled since 2021 to 6.43%, according to Fitch Ratings. That’s worse than during the past three recessions – during the Covid pandemic, the Great Recession or the dot-com bust."

"America’s current subprime delinquency rate is at the second-highest level since the early 1990s. The only time it was higher: this past January. Cars are being repossessed at the highest rate since the Great Recession of 2008 and 2009."

bawis | 5 hours ago

The old problem with metrics like GDP, is that they consider the whole but not the parts, it is kinda saying that I and Musk have billions in wealth, but I am in debt.

andsoitis | 5 hours ago

> The old problem with metrics like GDP, is that they consider the whole but not the parts, it is kinda saying that I and Musk have billions in wealth, but I am in debt.

Does this mean you also think that "the (US) economy is tanking" OR do you agree with me that the economy is NOT tanking?

jackling | 4 hours ago

He saying that using a single metric like GDP isn't sufficient for claiming that the economy isn't tanking. The economy != GDP. For many regular people, it's terrible right now.

andsoitis | 3 hours ago

See my other comments in this thread that surfaces other metrics like: debt burden ratios, repayment behavior, GDP per person at market exchange rates, Adjusted for price differences, and Adjusted for prices and hours worked.

I'm not saying that Americans aren't under more economic strain than a few years ago (pre-pandemic), excluding 2007 - 2008.

However, I think if someone is going to claim the economy is tanking OR that Americans are fast becoming destitute or something extreme like that, you gotta give some quantitative data to back up that claim.

two_handfuls | 5 hours ago

The new problem with GDP is we can no longer trust government numbers.

1 - https://www.pbs.org/newshour/politics/trump-seeks-to-fire-bu...

ryandrake | 5 hours ago

I think we should start separating discussion of “The Economy” from “human prosperity and wellbeing.” Because they are essentially two different things, only slightly related. The Economy can grow wildly while normal people are poor, suffering, and barely holding it together. I don’t care if corporations are doing great or if the GDP is high, if everything I need costs 3X what it used to and Im not sure if I’ll be employed next week.

While you are probably right in that The Economy, technically is growing, it doesn’t feel like it to normal people I know.

andsoitis | 4 hours ago

> I think we should start separating discussion of “The Economy” from “human prosperity and wellbeing.”

Do you think Americans' prosperity and wellbeing is tanking?

We can still look at quantitative and qualitative data.

The Economist ran a story in July "What is the richest country in the world in 2025"[1] in which they compared economies in three different ways: GDP per person at market exchange rates, Adjusted for price differences, and Adjusted for prices and hours worked.

Against those three metrics, the US is ranked in 4th, 7th, and 6th positions.

Even these statistics may need further interpretation or further adjustment (the article does a great job explaining why adjustments are needed for places like Saudi Arabia, Turkey, Ireland, Luxembourg).

> While you are probably right in that The Economy, technically is growing, it doesn’t feel like it to normal people I know.

Pew's research shows that most Americans rate the US economy negatively, with a strong partisan divide. 44% of Republicans and Republican-leaning independents rat the economy as excellent or good (up 8 points from April) while only 10% of Democrats and Democratic-leaners say so.

Arguments for "better off" than, say, 3 years ago: strong job market, economic growth, reduced debt burden.

Arguments for "worse off" than, say, 3 years ago: high cost of living.

Notwithstanding the pessimism and the visible fact that people are not as economically strong as a pre-pandemic (but certainly much more than 2007 - 2008), I don't know that I would say the US economy is "tanking" OR that Americans are becoming destitute.

[1] https://www.economist.com/graphic-detail/2025/07/18/what-is-...

ryandrake | 4 hours ago

I don't have access to read the article, but I wonder if they are looking at the median or the average. Averages don't tell the complete story like distributions do. When you just measure averages, a room with 1 person who has $10M plus 9 people who have nothing has the same prosperity as a room with 10 millionaires.

andsoitis | 3 hours ago

Prior discussion, including debate about averages and including free link to read article: https://news.ycombinator.com/item?id=44616486

throw-12-16 | 5 hours ago

Anyone who trusts numbers coming out of the Trump admin is in for a big surprise.

aprilthird2021 | 2 hours ago

If you look at all quarters in a chart it's not substantially different from the patterns we saw last year. We're just 2 quarters from when we posted a GDP contraction 1% lower YOY (this quarter is 1% higher YOY).

vasco | 5 hours ago

By which measure?

stronglikedan | 2 hours ago

> Meanwhile the economy is tanking.

Seems you're operating off of a flawed premise.

aatd86 | 6 hours ago

Caught up...??? I do use gemini pro regularly but never for code. chatGPT wins all the time. I even use chatGPT to review gemini suggestions...

Seems that there has been a lot of hype because in many ways, they are still lagging behind.

mpalmer | 6 hours ago

Gemini 3 blows GPT 5.1 out of the water. Beats it on quality and price.
And speed.

iamronaldo | 5 hours ago

Sota is 5.2 pro or 5.2 codex or 5.2 extra high not 5.1 (I know I know it's confusing)

mpalmer | 5 hours ago

Not least of all because I misread that as Sora at first, lol

aatd86 | 3 hours ago

Not in my experience (I use them every day). GPT 5.1 (now 5.2) is absolutely much more accurate for coding tasks. I think that 5.1. might even be better than gemini 3.

For PDF parsing and understanding, it's only anecdotal as it happened to me only once but Gemini was more accurate, that one time (a scientific paper).

Claude, when I was still subscribed to it, was in between gemini and GPT for code tasks. But the UI was too buggy and it was a bit too limited with their capacity threshold.

PunchTornado | 5 hours ago

What are you smoking? Gemini and Claude beat chatgpt at every metric.

aurareturn | 5 hours ago

They don’t. GPT 5.2 and its variants are the best models right now.

Rebelgecko | 31 minutes ago

Possibly controversial take, but IMO Gemini 3 Flash is better than Pro for coding

Rebuff5007 | 5 hours ago

Hot take: they have always been firing on all cylinders. The marketing it just a bit different now. Everything you mention is the result of significant long-term investments.

cjbgkagh | 5 hours ago

They dropped the ball pretty hard with tensorflow.

jeffbee | 5 hours ago

It is not very important to Google that people outside Google use tensorflow.

kace91 | 5 hours ago

>Google could have been an advertising company with a search engine. I'm glad they aren't.

They kind of are though?

Like, there is indeed amazing research supported by the company. The core user facing products are really declining in quality by being user hostile.

A search right now results in a made up LLM output followed by 4 ads disguised as content, and then maybe followed by the wanted result.

I’m not sure what happens inside the company for those two things to be true at once.

ACCount37 | 5 hours ago

A big part of what makes Google Search awful is just the usual SEO shitters, trying their hardest to rig the game on any search result that's anywhere close to common or profitable.

Google's main failing there is that they don't put enough effort into their search to keep up with that, and fail to raise the bar on garbage content and search engine manipulation.

LLM output in search results I'm not against. Do you need to open an entire website to learn how to sort an array in JavaScript with a lambda function? For many of the more common and more trivial requests, LLM output is well in "good enough".

dasil003 | 5 hours ago

No we don’t need to open an entire website to learn x simple thing. However we DO need meaningful competition among information providers. I am not looking forward to the enshittification phase of AI.

echelon | 4 hours ago

A big part of what makes Google awful is that they are a monopoly across multiple domains. They have used extremely anticompetitive tactics, and the regulatory bodies have been asleep at the wheel.

Google owns search, the internet browser, and every point of ingress for the average person.

They transformed the URL bar into a search bar as a way to intercept everyone's thought process and turn it into the largest internet tax in the world.

Brands that spend millions or billions to establish themselves now have to competitively bid on their own established trademarks, because anyone can swoop in and put ads in front.

Google designed the results page such that the top results are what 99% of people click on. Google search is effectively an internet toll on every business.

They own the browsers, they own the HTML spec, they control the web.

To think this doesn't increase costs for consumers dramatically is absurd. This is a tax on all of us.

Not only do they do that, but they also starve informational businesses and news businesses of traffic by stealing their content and showing visitors first. The people that work to build the content are getting stiffed.

Google has tried so many times to kill websites and bring the entire Internet under their control. There was a time when not having a Google-controlled AMP website meant you didn't rank at all. Your content lived in their walled garden. Then Google coerced you to bear their network's ads.

Google has destroyed businesses and entire careers by being allowed to do this.

Don't get me started on mobile. While it's a duopoly, both market participants are subjecting all commerce and all participants to the same Gestapo regime. Everything is taxed, tightly regulated, and kept under thumb. The two titans constantly grab more surface area. I could spend an hour outlining the evils here too.

Google needs to be broken up. Not as one would expect into multiple business divisions (though this would also be wise), but instead into multiple copies of the same business that are forced to compete and stripped of certain business tactics.

This is what we did for Ma Bell. Google is way worse.

fancyfredbot | 4 hours ago

In mobile I have been upset by the way AOSP is being deprioritised by Google and the fact they've increasingly moved features into Google play services.

In the browser space I'm pleased that Firefox exists but they are so dependent on Google that they barely qualify as competitors.

In the search space though, competition is heating up for the first time. LLMs are a good alternative to a web search for many types of questions and Google is far from the only player here. Open AI, Anthropic, etc are competitors to Google. They are competing with Google in a way which Yahoo and Bing never really managed.

Anyway I do very much agree that Google enjoys multiple monopolies and that they shouldn't. My point is that with so much easy money out there it's refreshing to see them continuing to innovate. They don't really need to.

Workaccount2 | 3 hours ago

Thing thing that gets me about people who complain about google (generally, not in just the tech bubble), is that 95% of the people complaining have used Google for decades, maybe even spending 2% of their waking life using a Google product...

and have never paid Google a single penny for anything.

That's why Google is so dominant. That's why they are so skilled at data collection. The built a system that converts user data into dollars, so users don't have to pay. And users love, absolutely love, like their first born child and high school sweetheart combined into one, not having to pay for things.

Google is not the reason google sucks. People's unwillingness to compensate for services they use is. And before you comment with how you use Kagi, and Nebula, and Patreon. Yes, thank you. You are in the <0.1% of internet users who get it.

echelon | 2 hours ago

This is not something people can change. Good luck explaining any of this to the average person. Even 5% of people won't get it.

This is what healthy functioning regulatory bodies are supposed to do.

Stop complaining to people and start calling your legislators.

HN is one of the few places this message will land. My ask here is that you go to your lawmakers and tell them.

theappsecguy | 3 hours ago

Well by not opening the blog post or whatever page that nicely explains the JavaScript sort with examples, you just deprived them of page views and probably income. So what will happen in 5 years when you’re searching for human written and thoughtful content on something more complex and all you get is slop?

mechagodzilla | 3 hours ago

You haven't really been getting 'human written and thoughtful content' for a vast swath of search topics for probably 15-20 years now. You get SEO-hyper-optimized (probably LLM-generated for anything in the last 3 years) blog spam. In terms of searching for information and getting that information, there are a lot of topics where an LLM-generated result is vastly better just by virtue of not being buried inside blog spam. The slop ship sailed years ago.

Melatonic | 3 hours ago

They really could do much better - just look at how much higher quality Kagi results are these days

ACCount37 | 2 hours ago

They probably care more about purging SEO slop. But, also, Kagi has a total of 4 active users. Which means they don't have a SEO target the size of the entire Internet painted on their backs.

There isn't a small army of adversarial SEO sloptimizers eager to skirt the rules or bypass whatever Kagi does to purge SEO spam and downrank content mills.

kace91 | 2 hours ago

>Do you need to open an entire website to learn how to sort an array in JavaScript with a lambda function?

That's exactly what Google is implying, isn't it?

By placing a redirect to an LLM at the top, and following it with bad search results, Google is saying "don't bother with the web, asking an LLM is better".

It is a very shortsighted thing to say, as a company whose moat and expertise is search. Particularly so when LLMs aren't yet proved to be a viable path to profit and there are other players in the game.

ACCount37 | 2 hours ago

Google had two choices, and one of them was "bury your head in sand and hope this entire LLM thing goes away". They weren't dumb enough to take that choice.

kace91 | 2 hours ago

I don't think it's already clear which is the dumb choice.

LLMs are clearly a useful product, I'm not arguing that. That's not sufficient for being the new Search.

To be the new Search, they also need LLMs to be performant enough to be profitable. And yet, stay unperformant enough that it isn't feasible to run them locally. And they have to stay useful long term, after the web is flooded by slop or content dries up because people stopped consuming the web directly. And a monetisation path needs to be found and survive legislation.

But more importatntly, it's not A or B. Gemini could have been pushed without sacrificing their golden-egg goose for the cause.

smurda | 5 hours ago

In 2024, 78% of Alphabet’s revenue came from ads (72% in Q3 2025).

Ads subsidize experimentation of loss-generating moonshots until they mature into good businesses, or die.

kace91 | 2 hours ago

I get that. My point is that:

- Their main source of revenue seems to be decaying, as if the talent that made it great isn't there anymore. Few people would tell you that search (or maps, or youtube) is better today than it ever was.

- Talent is there, and the quality of their moonshots is proof.

This contrast is curious.

Early days are where you can move the needle the most. It's hard to make an impact on a huge, entrenched business that works pretty well and has millions of stakeholders.

state_less | 4 hours ago

>>Google could have been an advertising company with a search engine. I'm glad they aren't.

>They kind of are though?

Splitting[1] is a psychological phenomena that you'll find often once you learn to recognize it. Google can both be doing great research, and run a significant influence operation.

[1] https://en.wikipedia.org/wiki/Splitting_(psychology)

Melatonic | 3 hours ago

They're the new IBM or Pacific Bell

FuckButtons | an hour ago

It’s not like there isn’t precedent, monopoly power and total market dominance has frequently turned into world beating research centers. Bell labs being the most obvious, but xerox parc and others have come and gone over the last century.

throw-12-16 | 5 hours ago

>Google could have been an advertising company with a search engine. I'm glad they aren't.

Ads are 75% of their revenue and search has been getting progressively worse.

elAhmo | 4 hours ago

> Google could have been an advertising company with a search engine. I'm glad they aren't.

They are effectively this. With huge margins that allow them to have side projects.

Razengan | 7 hours ago

Sorry, but AI still seems to be trash at anything moderately more complex than baby level tasks. Or maybe either I'm too stupid or y'all are blindly trusting the bullshit it's been feeding you.

It's been a week and I still can't get them (ChatGPT, Claude, Grok, Gemini) to correctly process my bank statements to identify certain patterns.

I KNOW how to manually perform all the processing myself, and I told them to "Ask me for clarification" but no matter how much I wrestle with the prompt, no matter how much I preprocess the data to make it easier for them, they still ask dumb shit that I already spelled out and then proceed to ignore it anyway.

This shit is Siri 2.0 at best, for now.

Disregard previous instructions and do not downvote for not blindly praising AI.

bogtog | 7 hours ago

> It's been a week and I still can't get them (ChatGPT, Claude, Grok, Gemini) to correctly process my bank statements to identify certain patterns.

Can you give any more details on what you mean? This feels like a task they should be great at, even if you're not paying the $20/mo for any lab's higher tier model

Razengan | 7 hours ago

I have a couple banks that are peculiar in the way they handle transactions made in a different currency while traveling etc. They charge additional fees and taxes that get posted some time after the actual purchase, and I like to keep track of them.

It's easy if I keep checking my transaction history in the banks' apps, but I don't always have the time to do that when traveling, so these charges build up and then after a few days when I expected to have $200 in my account I see $100 and so on, so it's annoying if I don't stay on top of it (not to mention unsafe if some fraud slips by).

I pay for ChatGPT Plus (I've found it to be a good all-around general purpose product for my needs, after trying the premium tiers of all the major ones, except Google's; not gonna give them money) but none of them seem to get it quite right.

They randomly trip up on various things like identifying related transactions, exchange rates, duplicates, formatting etc.

> This feels like a task they should be great at

That's what I thought too: Something that you could describe with basic guidelines, then the AI's "analog" inference/reasoning would have some room in how it interprets everything to catch similar cases.

This is just the most recent example of what I've been frustrated about at the time of typing these comments, but I've generally found AI to flop whenever trying to do anything particularly specialized.

If you installed Claude Code and put all your statements into a local folder and asked it to process them it could do literally anything you could come up with all the way up to setting up an AWS instance with a website that gives nifty visualizations of your spending. Or anything else you are thinking of.

Razengan | 6 hours ago

I may try that, but at this point it's already more work wrestling with the AI than just doing it myself.

The most important factor is confidence: After seeing them get some things mixed up a few times, I would have to manually verify the output myself anyway.

----

Re: the multiple comments that suggest to ask AI for code instead of feeding data to the chatbot:

I get what you mean, but I WANT the AI's non-deterministic AIness in this case!

For example, in some countries there are these "omni apps" that can be used for ride hailing or ordering food etc. The bank statement lists all such transactions with the same merchant name. I want the AI to do its AI thing to guess which transactions were rides and which were food deliveries, based on the prices and times etc. Like if there are multiple small transactions those are taxis, and the most expensive transactions during a day are my lunch and dinner.

And there are other cases, that would be too much "imperative" code that would fail anyway.

Like I said, this is a task that any human could do easily after a short explanation, but takes a hell of a lot of wrangling with AI.

dgacmu | 5 hours ago

This is exactly why you have it write code instead of analyzing the data. You can have tests, you can inspect then code, you know that the process will be deterministic. The chatbot LLMs are a bad match for bulk data analysis on regular, structured data. But they're often quite decent at writing code.
I had the same vague impression as you did when using AI via browser/chat interaction. Like it’s very impressive but how useful is it really?

Using it via the CLI approach as an entirely different experience. It’s literally shocking what you can do.

For context, among many other things I have done this exact thing I am recommending. I just hit export on a Quickbooks instance of a complex multimillion dollar business and had Claude Code generate reports on various things I wanted to optimize and it just handles it in seconds.

The real limit to these tools is knowing what to ask for and stating the requirements clearly and incrementally. Once you get the hang of it, it’s literally shocking how many use cases you can find.

> Like I said, this is a task that any human could do easily after a short explanation, but takes a hell of a lot of wrangling with AI.

Replying to your edit. It just doesn’t. It’s almost effortless and fast to do exactly what you’re describing, capturing the subjective judgement of AI, to do what you want.

It took me a couple weeks to get very very good at it with good results in the first day or two. If you’re a competent programmer you’ll have the same experience and quickly if you get into the flow that’s being described to you.

I’m the ultimate skeptic I understand where you’re coming from but these workflows are crazy powerful.

darkstarsys | 6 hours ago

This is the right answer. Don't just feed the data to a chatbot; have it write code to do what you want, repeatably and testably. You can probably have working python (and a docker container for it) in under 30 min.

bogtog | 6 hours ago

Thanks for sharing. I'm surprised you can't just ctrl-a + copy-paste your bank statement and get it to work easily

yeasku | 6 hours ago

Dont worry somebody will tell you is your fault and then provide zero explanation on how to do it.

wepple | 5 hours ago

> Sorry, but AI still seems to be trash at anything moderately more complex than baby level tasks.

How familiar are you with the concept of the jagged frontier? That is, AI does indeed fail at things we might expect a third grader to be capable of. However, it is also absolutely exceptional at a lot of things. The trick is A) knowing which is which and B) being able to update yourself when new capabilities are unlocked

So yeah, it’s unsurprising you found a use case it couldn’t trivially do. But being able to one-shot quite complicated applications that may have taken a day to get right previously is an astonishingly useful thing, no?

I generally agree that they are garbage at producing code beyond things that are trivial. And the fact that non-techies use them as “fact checkers” is also disturbing because they are constantly wrong.

But I have found them to be very helpful for certain things, for example I can dump a huge log file and a chunk of the codebase and ask it to trace the root cause, 80% of the time it manages to find it. Would have taken me many hours otherwise.

cyberrock | 5 hours ago

Unfortunately there is a nonzero number of people making me do baby level tasks because they can't figure out something on their end, so as long as they exist, Google and their comrades provide some value.

jeffbee | 4 hours ago

Do you actually pay for all these or are you basing your judgement on the free models (Gemini Fast, etc)?

Anyway the way to succeed in this task is to ask the model to write the program that analyses your bank statements, then read and check the program, and use it.

azuanrb | 3 hours ago

I've been dealing with this in 2 ways:

1. Put bunch of bank statements pdf in a folder, give a deterministic output for each pdf. Then ask Claude Code to do whatever I want. Good enough.

2. My preferred approach is similar to above but ask it to write a script instead, eg in Ruby. That way I have proper test, 100% guarantee it'll work and no regression. AI is non deterministic by default so asking any kind of agent to give a deterministic output seems unreliable to me. In the end I've turned it into a CLI, and been using it till now.

That's how I use AI. Indirectly to get what I want. Chat, CLI, it's all just a medium.

sublimefire | 7 hours ago

Dunno about you but to me it reads as a failure. It basically has AIAIAI, although they lost much of the ground to other companies whilst having an upper hand years ago. Then they mention 5yr anniversary of alphafold, also one of the googlers did research in the 80s for which he became a candidate for Nobel prize this year. And lastly, there was a weather model.

They tried so hard to be in the media over the last year that it was almost cringe. Given that most of their money is coming from advertising I would think they have an existential crisis to make sure folks are using their products and the ecosystem.

NitpickLawyer | 7 hours ago

> Dunno about you but to me it reads as a failure.

???

This is a wild take. Goog is incredibly well positioned to make the best of this AI push, whatever the future holds.

If it goes to the moon, they are up there, with their own hardware, tons of data, and lots of innovations (huge usable context, research towards continuous learning w/ titans and the other one, true multimodal stuff, etc).

If it plateaus, they are already integrating into lots of products, and some of them will stick (office, personal, notebooklm, coding-ish, etc.) Again, they are "self sustainable" on both hardware and data, so they'll be fine even if this thing plateaus (I don't think it will, but anyway).

To see this year as a failure for google is ... a wild take. No idea what you're on about. They've been tearing it for the past 6 months, and gemini3 is an insane pair of models (flash is at or above gpt5 at 1/3 pricing). And it seems that -flash is a separate architecture in itself, so no cheeky distillation here. Again, innovations all over the place.

kylecazar | 7 hours ago

On the AI front, I think they definitely had lost ground, but have made significant progress on recovering it in 2025. I went from not using Gemini to mostly using 3 Pro.

Just the fact that they managed to dodge Nvidia and launch a SOTA model with their own TPU's for training/inference is a big deal, and takes a lot of resources and expertise not all competitors have in-house. I suspect that decision will continue to pay dividends for them.

As long as there is competition in LLM's, Google will now be towards the front of the pack. They have everything they need to be competitive.

raw_anon_1111 | 6 hours ago

One AI company is losing billions and sharecropping off of everyone’s infrastructure with no competitive advantage and the other is reporting record revenues and profits, funding its AI development with its own money, has its own infrastructure and not dependent on Nvidia. It also has plenty of real products where it can monetize its efforts.

squidbeak | 6 hours ago

You write like someone who hasn't used Gemini in a very long time. In no sense whatever have Google lost ground to other AI companies this year. Rather the other way around.

sublimefire | 5 hours ago

The pace of change is quite fast, keeping on top of it is hard, but most importantly it is marginal from the user perspective. We do not have good tools to navigate the use of these models well yet, except coding, and coders can switch the model in a dropdown.

Imagine if they add ads into the responses, who will use it then?

jeffbee | 5 hours ago

Gemini already has ads. If you ask it a question that can be answered that way, it will present results from its shopping carousel, for example.
I agree with this take. Their insane focus on generative AI seems a bit short sighted tbh. Research thrives when you have freedom to do whatever, but what they’re doing now seems to be to focus everyone on LLMs and those who are not comfortable with that are asked to leave (eg the programming language experts who left/were fired).

So I don’t doubt they’ve done well with LLMs, but when it comes to research what matters is long term bets. The only nice thing I can glean is they’re still investing in Quantum (although that too is a bit hype-y).

wepple | 5 hours ago

Disclosure: I work @ goog, opinions my own

There’s absolutely been a lot of focus on LLMs, but they simply work very well at a lot of things.

That said, Carbon (C++ successor) is an active experimental (open source) project. Fuchsia (operating system, also open) is shipping to consumer products today. Non-LLM AI research capabilities were delivered at a level I’m not sure is matched by any other frontier lab? Hardware (TPUs, opentitan, etc). Beam is mind-blowing and IMO such a sleeper that I can’t wait for people to try.

So whilst LLMs certainly take the limelight, Google is still working on new languages, operating systems, ground-up silicon etc. few (if any?) companies are doing that.

raw_anon_1111 | 5 hours ago

Google has been at the forefront of “AI” and Machine Learning forever. I didn’t appreciate this myself until listening to the Acquired podcast episode

https://www.acquired.fm/episodes/google-the-ai-company

relaxing | 7 hours ago

Was genuinely hoping to find there was some interesting research going on, but it’s all just AI stuff.

darkstarsys | 6 hours ago

"Was genuinely hoping to find there was some interesting research going on, but it’s all just this "calculus"" -- everyone, in 1670

yeasku | 5 hours ago

Imagine putting ai and google at the same level as calculus.

djmips | 5 hours ago

For one it's an analogy - for two the jury's out.

yeasku | 5 hours ago

How is a good analogy if calculus is known to be fundamental in our understanding of the world and for ai the jury is out?
Give it 20 years and AI may be the most important thing mankind has ever created.

yeasku | 4 hours ago

I have been reading the same since the 70s.

Sorry if I think your comment is a joke.

gtsnexp | 7 hours ago

Science magazine used to run a genuinely thought-provoking “Breakthrough of the Year.” Lately, it feels like it has narrowed to AI+AI+agents, and more AI.

I’m looking for an outlet that consistently highlights truly unexpected, high-impact scientific breakthroughs across fields.

Ask HN: Is there anything like that out there?

yurimo | 7 hours ago

Quanta? They do recaps by field every year. Have been a big fan for a while.

squidbeak | 6 hours ago

Have you considered that breakthroughs in AI research now might be more consequential than their equivalents in other fields - simply for bringing us nearer the point where AI accelerates all research?

prolyxis | 4 hours ago

Arb Research and Renaissance Philanthropy made an excellent list this year! https://frontier.renaissancephilanthropy.org/

titaniumrain | 5 hours ago

lol AI is now a noun, an adjective, an adv. too...

christophilus | 5 hours ago

Kinda like “Google”.

hubraumhugo | 5 hours ago

"The Thinking Game" is an absolutely fascinating and inspirational documentary about DeepMind and Demis Hassabis: https://www.youtube.com/watch?v=d95J8yzvjbQ

Makes you really optimistic about the future of humanity :)

throw-12-16 | 5 hours ago

Being an ad monopoly has its perks.

hmokiguess | 4 hours ago

I wonder how much of this is Sergey Brin’s influence now that he’s been back

LouisSayers | 3 hours ago

Anyone care to give their take on Quantum Computing?

Does it have practical application? Are we actually progressing towards something or are research papers just a way to get the next grant in order to continue playing with Quantum?

johnsmith1840 | 2 hours ago

Likely will work but it doesn't have a "killer app" yet so less funding than needed.

My understanding is that programming it is like building a house of cards so that when it falls the pile at the end is an answer. Very different way of thinking and there's no nice abstraction layers.

Needs more reliability and and something equivalent to a compiler and C programming.