I was recently (vibe)-coding some games with my kid, and we wanted some basic text-to-speech functionality. We tested Google's Gemini models in-browser, and they worked great, so we figured we'd add them to the app. Some fun learnings:
1. You can access those models via three APIs: the Gemini API (which it turns out is only for prototyping and returned errors 30% of the time), the Vertex API (much more stable but lacking in some functionality), and the TTS API (which performed very poorly despite offering the same models). They also have separate keys (at least, Gemini vs Vertex).
2. Each of those APIs supports different parameters (things like language, whether you can pass a style prompt separate from the words you want spoken, etc). None of them offered the full combination we wanted.
3. To learn this, you have to spend a couple hours reading API docs, or alternatively, just have Claude Code read the docs then try all different combinations and figure out what works and what doesn't (with the added risk that it might hallucinate something).
- The models perform differently when called via the API vs in the Gemini UI.
- The Gemini API will randomly fail about 1% of the time, retry logic is basically mandatory.
- API performance is heavily influenced by the whims of the Google we've observed spreads between 30 seconds and 4 minutes for the same query depending on how Google is feeling that day.
> The Gemini API will randomly fail about 1% of the time, retry logic is basically mandatory.
That is sadly true across the board for AI inference API providers. OpenAI and Anthropic API stability usually suffers around launch events. Azure OpenAI/Foundry serving regularly has 500 errors for certain time periods.
For any production feature with high uptime guarantees I would right now strongly advise for picking a model you can get from multiple providers and having failover between clouds.
Yeah at $WORK we use various LLM APIs to analyze text; it's not heavy usage in terms of tokens but maybe 10K calls per day. We've found that response times vary a lot, sometimes going over a minute for simple tasks, and random fails happen. Retry logic is definitely mandatory, and it's good to have multiple providers ready. We're abstracting calls across three different APIs (openai, gemini and mistral, btw we're getting pretty good results with mistral!) so we can switch workloads quickly if needed.
I've been impressed by ollama running locally for my work, involving grouping short text snippets by semantic meaning, using embeddings, as well as summarization tasks. Depending on your needs, a local GPU can sometimes beat the cloud. (I get no failures and consistent response times with no extra bill.) Obviously YMMV, and not ideal for scaling up unless you love hardware.
I had this last night from flash lite! My results were interspersed with random snippets of legible, non-gibberish English language. It was like my results had got jumbled with somenone else's.
I get this a lot too, have made most of the Gemini models essentially unusable for agent-esque tasks. I tested with 2.5 pro and it still sometimes devolved into random gibberish pretty frequently.
Also, usage and billing takes a DAY to update. On top of that, there are no billing caps or credit-based billing. They put the entire burden on users not to ensure that they don't have a mega bill.
4. If you read about a new Gemini model, you might want to use it - but are you using @google/genai, @google/generative-ai (wow finally deprecated) or @google-ai/generativelanguage? Silly mistake, but when nano banana dropped it was highly confusing image gen was available only through one of these.
5. Gemini supports video! But that video first has to be uploaded to "Google GenAI Drive" which will then splices it into 1 FPS images and feeds it to the LLM. No option to improve the FPS, so if you want anything properly done, you'll have to splice it yourself and upload it to generativelanguage.googleapis.com which is only accessible using their GenAI SDK. Don't ask which one, I'm still not sure.
6. Nice, it works. Let's try using live video. Open the docs, you get it mentioned a bunch of times but 0 documentation on how to actually do it. Only suggestions for using 3rd party services. When you actually find it in the docs, it says
"To see an example of how to use the Live API in a streaming audio and video format, run the "Live API - Get Started" file in the cookbooks repository".
Oh well, time to read badly written python.
7. How about we try generating a video - open up AI studio, see only Veo 2 available from the video models. But, open up "Build" section, and I can have Gemini 3 build me a video generation tool that will use Veo 3 via API by clicking on the example. But wait why cant we use Veo 3 in the AI studio with the same API key?
8. Every Veo 3 extended video has absolutely garbled sound and there is nothing you can do about it, or maybe there is, but by this point I'm out of willpower to chase down edgy edge cases in their docs.
9. Let's just mention one semi-related thing - some things in the Cloud come with default policies that are just absurdly limiting, which means you have to create a resource/account, update policies related to whatever you want to do, which then tells you these are _old policies_ and you want to edit new ones, but those are impossible to properly find.
10. Now that we've setup our accounts, our AI tooling, our permissions, we write the code which takes less than all of the previous actions in the list. Now, you want to test it on Android? Well, you can:
- A. Test it with your account by signing in into emulators, be it local or cloud, manually, which means passing 2FA every time if you want to automate this and constantly risking your account security/ban.
- B. Create a google account for testing which you will use, add it to Licensed Testers on the play store, invite it to internal testers, wait for 24-48 hours to be able to use it, then if you try to automate testing, struggle with having to mock a whole Google Account login process which every time uses some non-deterministic logic to show a random pop-up. Then, do the same thing for the purchase process, ending up with a giant script of clicking through the options
11. Congratulations, you made it this far and are able to deploy your app to Beta. Now, find 12 testers to actively use your app for free, continuously for 14 days to prove its not a bad app.
At this point, Google is actively preventing you from shipping at every step, causing more and more issues the deeper down the stack you go.
Trying to implement their gRPC api from their specs and protobufs for Live is an exercise in immense frustration and futility. I wanted to call it from Elixir, even with our strong AI I wasted days then gave up.
Has anyone had success getting Claude to write it's own Claude.md file? It should be able to deduce rules by looking at the code, documentation, and PR comments.
The main failure state I find is that Claude wants to write an incredibly verbose Claude.md, but if I instruct it "one sentence per topic, be concise" it usually does a good job.
That said, a lot of what it can deduce by looking at the code is exactly what you shouldn't include, since it will usually deduce that stuff just by interacting with the code base. Claude doesn't seem good at that.
An example of both overly-verbose and unnecessary:
### 1. Identify the Working Directory
When a user asks you to work on something:
1. *Check which project* they're referring to
2. *Change to that directory* explicitly if needed
3. *Stay in that directory* for file operations
```bash
# Example: Working on ProjectAlpha
cd /home/user/code/ProjectAlpha
```
(The one sentence version is "Each project has a subfolder; use pwd to make sure you're in the right directory", and the ideal version is probably just letting it occasionally spend 60 seconds confused, until it remembers pwd exists)
If you have any substantial codebase, it will write a massive file unless you explicitly tell it not to. It also will try and make updates, including garbage like historical or transitional changes, project status, etc...
I think most people who use Claude regularly have probably come to the same conclusions as the article. A few bits of high-level info, some behavior stuff, and pointers to actual docs. Load docs as-needed, either by prompt or by skill. Work through lists and constantly update status so you can clear context and pick up where you left off. Any other approach eats too much context.
If you have a complex feature that would require ingesting too many large docs, you can ask Claude to determine exactly what it needs to build the appropriate context for that feature and save that to a context doc that you load at the beginning of each session.
An alternative is to view the AI agent as a new developer on your team. If existing guidance + one-shot doesn't work, revisit the documentation and guidance (ie dotMD file), see what's missing, improve it, and try again. Like telling a new engineer "actually, here is how we do this thing". The engineer learns and next time gets it right.
I don't do MCPs much because of effort and security risks. But I find the loop above really effective. The alternative (one-shot or ignore) would be like hiring someone, then if they get it wrong, telling them "I'll do it myself" (or firing them)... But to each his own (and yes, AI are not human).
I don't think you can say it learns - and that is part of the issue. Time mentoring a new colleague is well spent making the colleague grow professionally.
Time hand-holding an AI agent is wasted when all you guidance inevitably falls out of the context window and it start making the same mistakes again.
> you guidance inevitably falls out of the context window
Yes, for a pure LLM session.
But using GitHub Copilot, the agent server picks out what it thinks is your most important rules, and inserts them as part of the context before running each new prompt.
So it becomes a sort of "long-term memory" outside of the context window.
You can also write your own copilot-instructions.md which is also inserted into the context window.
On the contrary; stubborn refusal to anthropomorphize LLMs is where the frustration comes from. To a first approximation, the models are like little people on a chip; the success and failure modes are the same as with talking to people.
If you look, all the good advice and guidelines for LLMs are effectively the same as for human employees - clarity of communication, sufficient context, not distracting with bullshit, information hygiene, managing trust. There are deep reasons for that, and as a rule of thumb, treating LLMs like naive savants gives reliable intuitions for what works, and what doesn't.
I treat LLMs as statistics driven compression of knowledge and problem solving patterns.
If you treat it as such it is all understandable where they might fail and where you might have to guide them.
Also treat it as something that during training has been biased to produce immediate impressive results. This is why it bundles everything into single files, try catch patterns where catch will return mock data to show impressive one shot demo.
So the above you have to actively fight against, to make them prioritise scalability of the codebase and solutions.
Honestly it sounds about right: at the end of the day, most companies will always be an interesting UI and workflow around some commodity tech, but, that's valuable. Not all of it may be defensible, but still valuable.
At this point, I think all of the big tech companies have had some accusations of them acting unethically, but usually, the accusations are around them acting anticompetitively or issues around privacy.
Meta (and social media more broadly) are the only case where we have (in my opinion) substantiated allegations of a company being aware of a large, negative impact on society (mental wellness, of teens no less), and still prioritizing growth and profit. The mix is usually: grow at all costs mindset, being "data-driven", optimizing for engagement/addiction, and monetizing via ads. The center of gravity of this has all been Meta (and social media), but that thinking has permeated lots of other tech as well.
It's a well worn playbook by now. But Meta seems to be the only one where we now have proof of internal research being scuttled for showing the inconvenient truth.
True, but there haven't even been publicly known internal research attempts at for example YouTube/Google about the content they are pushing and probably more importantly the ads they keep pushing into people's faces. I bet FB/Meta are kicking themselves now, for even running such internal research in the first place.
My point is, that all of these big tech giants will find, that they are a harmful cancer to society, at least in parts. Which is probably why they don't even "research" it. This way they can continue to act oblivious to this fact.
> I bet FB/Meta are kicking themselves now, for even running such internal research in the first place.
100%. This is what people miss in this thread when they're talking about seeing to punish companies who knowingly harm society. All that is going to do is discourage companies from ever seeking to evaluate the effects that they're having.
The tobacco industry also did that, but in many ways it also seems different, because where tobacco was something that has existed for millennia and was a scourge introduced to the world by the tribes of the “new world”; Facebook was a primary player in creating the whole social media space, something that effectively did not exist in the predatory and malignant manner that it was used for to create a digital panopticon, or more accurately and way worse, where your participation is required for a certain kind of success.
Social media is abusive and utterly psychotic and narcissistic, because that is the type of people who created it using basic psychological abuse and submission tactics. Banks, casinos, games, hollywood/TV, news/politics, social media, contemporary academia and religion, etc.; they all function on being endorphin dealers/dispensers.
What do you think the social effects of large scale advertising are? The whole point is to create false demand essentially driving discontent. I've no idea if Google et al have ever done a formal internal study on the consequences, but it's not hard to predict what the result would be.
The internet can provide an immense amount of good for society, but if we net it on overall impact, I suspect that the internet has overall had a severely negative impact on society. And this effect is being magnified by companies who certainly know that what they're doing is socially detrimental, but they're making tons of money doing it.
I agree false demand effects exist. But sometimes ads tell you about products which genuinely improve your life. Or just tell you "this company is willing to spend a lot on ads, they're not just a fly-by-night operation".
One hypothesis for why Africa is underdeveloped is they have too many inefficient mom-and-pop businesses selling uneven-quality products, and not enough major brands working to build strong reputations and exploit economies of scale.
> But sometimes ads tell you about products which genuinely improve your life.
I’d argue that life improvement is so small it’s not worth the damage of false demand. I can maybe think of one product that I saw a random ad for that I actually still use today. I’d say >90% of products being advertised these days are pointless garbage or actually net negative.
Advertising is cancer for the mind and our society severely underestimates the harm it’s done.
The positive benefits in education, science research and logistics are hard to understate. Mass advertising existed before the internet. Can you be more explicit about which downsides you thibk the additional mass advertising on the internet caused that can come anywhere close to the immeasurable benefits provided by the internet?
I'm somewhat unsurprised that my off the cuff hypothesis has been tested, and is indeed likely accurate. [1] Advertising literally makes people dissatisfied with their lives. And it's extremely easy to see the causal relationship for why this is. Companies like Google are certainly 100% aware of this. And saying that advertising existed before the internet is somewhat flippant. Obviously it did but the scale has increased so dramatically much that it's reaching the point of absurdity.
And a practical point on this topic is that the benefits of the internet are, in practice, fringe, even if freely available to everyone. For instance now there are free classes from most of all top universities online, on just about every topic, that people can enroll and participate in. There are literally 0 barriers to receiving a free premium quality education. Yet the number of people that participate in this is negligible and overwhelmingly composed of people that would have had no less success even prior to the internet.
By contrast the negatives are extremely widespread on both an individual and social level. As my post count should demonstrate, I love the internet. And obviously this site is just one small segment of all the things I do on the internet. In fact my current living would be impossible without it. Yet if I had the choice of pushing a button that would send humanity on a trajectory where we sidestep (or move along from) the internet, I wouldn't hesitate in the slightest to push it.
> I'm somewhat unsurprised that my off the cuff hypothesis has been tested, and is indeed likely accurate.
That study is a correlation with self reported satisfaction. The effect size is that a doubling of ad spend results in a 3% change in satisfaction. I struggled to find good numbers but it appears as if ad spending in the USA has been a more or less constant percentage of GDP growth.
Thus the only real conclusion you can draw from your argument is that any increase in unhappiness caused by the internet was caused by its associated GDP growth increasing ad spend per capita.
Personally, I do think advertising has become more damaging precisely due to the internet but good luck proving it.
> And a practical point on this topic is that the benefits of the internet are, in practice, fringe, even if freely available to everyone
Ok, nevermind. I can't take anything else you say seriously when you call the ways the the internet has improved people's lives "fringe". I take it you never tried to take a bus pre-internet? Drive a car across the cohntry? Or lookup any information? The internet's effects on people is so far from fringe that it has seeped into almost everything we do at a fundamental level. Perhaps because of that you can't see it.
As the article mentions a 3% drop in life satisfaction is "about half the drop in life satisfaction you’d see in a person who had gotten divorced or about one-third the drop you’d see in someone who’d become unemployed." And advertising spending is increasingly exponentially. Good numbers on ad spend are available here [1], as that's the source they used (the exact date).
Ad spending was estimated at growing around 14% per year. In current times it's settled around 5-10% per year, but of course keep in mind that that's a compounding value. So a doubling isn't every 10-20 years but every ~7-14. And furthermore in their study they were able to demonstrate that shifts in happiness followed even local shifts in advertising. So when advertisers scaled back for various reasons, life satisfaction increased, and then began diminishing as the advertising returned.
Mass advertising will likely be the tobacco of our time once you consider the knock-on effects of societies full of individuals being made intentionally discontented.
You have yet to offer a single coherent counter-argument or explanation. Rather you are wildly flailing with seemingly endless weak claims that mean nothing. Like here if I look it up and the ad spend was indeed growing at somewhere in the ballpark of their numbers (since their predictions are obviously going to align with trends), are you going to change your opinion? Obviously not, no more than you're going to change your opinion after realizing that a 3% decline in satisfaction was indeed a lot, or that the study did indeed make some significant effort to demonstrate causality.
Or now if you learn that economies, especially European, grow dramatically slower [1] than you seem to think, are you going to change your opinion? No, because once again you're not making a logical or rationale argument whatsoever, but behaving like somebody on the defensive desperately trying to argue against something but having no foundation beyond the dislike of a seemingly inescapable conclusion.
That is 3% if you double ad spend. You've failed to demonstrate that the internet increased ad spend at all, let alone came anywhere near doubling it.
> Or now if you learn that economies, especially European, grow dramatically slower
I just was telling you what your own source claimed, not endorsing those claims...
> behaving like somebody on the defensive desperately trying to argue against something but having no foundation beyond the dislike of a seemingly inescapable conclusion.
Perhaps you should re-read the comment you just left...
So what is your argument supposed to be? That people don't see anymore ads than they used to, or that the internet is having no impact on the number of ads people are seeing? See the problem with how you're arguing? Obviously you don't believe these things, but it seems that's what you've apparently found yourself trying to argue.
Since you have forgotten what we are discussing, you made this claim:
> The internet can provide an immense amount of good for society, but if we net it on overall impact, I suspect that the internet has overall had a severely negative impact on society.
I am disputing the claim the net effect of the Internet on society has been severely negative.
> That people don't see anymore ads than they used to,
This seems hard to measure and the results would depend on how you define "seeing more ads". The result is irrelevant to your argument though because the one study you cited looks at ad spend, not "the number of ads people are seeing" so you can't generalize.
You've staked out a very strong claim here but have done a very poor job of backing it up.
It also just occurred to me in our discussion that you may have missed the fundamental point. The study used ad spend as a proxy for ads viewed. This is because measuring exactly how many ads people see, let alone over time, is impossible to measure, but it's undoubtedly increasing with a sharp exponential, largely thanks to the internet.
Ad spend works as a passable proxy for it, and is likely understating the impact in modern times since advertising has become cheaper than ever, again thanks to the internet. So ad spend is going up at the same time that the number of ads per spend is increasing, at a rate substantially faster than during their study period.
> The study used ad spend as a proxy for ads viewed
Yes, clearly that was the intent. That doesn't mean you can generalize from your proxy to the thing you have a hard time measuring.
> So ad spend is going up at the same time that the number of ads per spend is increasing, at a rate substantially faster than during their study period.
It seems like you are just making up facts.
The fact is that global spending on advertising is a fairly steady percentage of the GPD. It does occilate a little up and down, but the "exponential" growth you are talking about is merely the exponential growth of GDP.
Thus your argument that internet has increased ad spend and thus decreased happiness is false.
Addionaly, while this is also hard to quantify, the trends for cost per ad view does seem to be moving in the opposite direction. The cost of an impression seems to be moving upwards, not downwards.
The study started in 1980 and carried on to 2013. So they started in an era where you had a mixture of extremely expensive (television) or extremely low reach (local newspaper classifieds) ads. In the internet era you have high reach low cost options such that for less than $100, like $30 adjusted to 1980 dollars, you can hit tens of thousands of people. And there are vastly more people advertising precisely because of that.
Your GDP thing is a complete red herring. GDP has no relevance on the meaning of increases in advertising. Finally here [1] is a graph of global advertising spend over the years. Yes it is increasing exponentially, and obviously so are the number of ads people are seeing.
> In the internet era you have high reach low cost options such that for less than $100, like $30 adjusted to 1980 dollars, you can hit tens of thousands of people.
You could buy a classified with the same or better reach and same price in 1980.
> Your GDP thing is a complete red herring. GDP has no relevance on the meaning of increases in advertising.
You've claimed that internet cause ad spending to grow. Ad spending only grew because the GDP grew so you can't blame the internet for it.
> Yes it is increasing exponentially
Again, because GDP is increasing exponentially.
Also your graph doesn't show total spend, only the breakdown by sector.
Obviously GDP growth does not magically make spending on something increase.
And similarly obviously classifieds didn't have even remotely near the same reach. Classifieds were an opt-in categorized system that was delivered to a small subset of classified viewers that was, itself, a small subset of all newspaper subscribers in society. And you're paying for small lines of text delivered to this opt-in audience as opposed to as opposed to graphical images, or even videos, imposed on people against their will.
> Obviously GDP growth does not magically make spending on something increase.
Its not magic, but spending on certain types of things do increase with GDP. Insisting that they don't, despite the evidence that they do historically because it is inconvenient for your argument isn't convincing.
> And similarly obviously classifieds didn't have even remotely near the same reach.
You are claiming that a classified in 1970 would not routinely reach tens of thousands of people?
> as opposed to graphical images, or even videos, imposed on people against their will
The nature of advertising has changes and I don't really doubt it has been made more harmful. What I don't see is a line between that and your fantastical claim that the internet is somehow a net negative.
Once again GDP has nothing to do with anything here. Whether it was higher, lesser, or whatever else does not matter. Advertising spending increased exponentially only because exponentially more money was being spent on advertising. How exactly that relates to GDP has no relevance, whatsoever.
In 1980 (this study didn't even cover 1970), there were something like 1750 newspapers for something like 60 million readers. So the subscribers per paper was already (relatively) very low. And with classifieds you get a very tiny subset of that that opt-in to reading the classifieds, and further opt-in to the section you are listed in. It should be beyond obvious that this has basically no relationship whatsoever to modern advertising.
> How exactly that relates to GDP has no relevance, whatsoever.
You are claiming that the increased spending on advertising is due to the internet. The relationship between GDP and ad spending pretty clearly indicates that the internet was not the cause of increased ad spend, except in as much as it increased GDP.
If the internet has any effect on ad spend growth, you would have seen the growth rates of GDP and ad spend diverge with the birth of the internet.
To see how clearly wrong you are consider this, part of GDP growth is population growth. You are claiming that population growth is irrelevant to ad spend growth?
> It should be beyond obvious that this has basically no relationship whatsoever to modern advertising.
Even your own numbers pretty clearly indicate that the pricing can be equivalent, which was my point.
GDP is an abstract measurement of other things that, itself, has no causal effect on anything. Yes, population can increase while GDP decreases. And spending on things also does not just naturally increase. It increases if, and only if, there is more spending on it. And no, classifieds are not like modern advertising and you clinging onto this reflects that you are not arguing in good faith.
PG/VG base is exactly the same stuff that has been used in foggers/hazers for decades. If there were negative health effects associated with the stuff, we'd have spotted it long ago. As for nicotine, well, it's the same stuff as in cigarettes, we know about its effects again thanks to decades of research.
The only thing left is questionable flavoring agents and dodgy shops with THC oil vapes (although that kind of contamination is now known and it's been ages since I last heard anything).
>PG/VG base is exactly the same stuff that has been used in foggers/hazers for decades. If there were negative health effects associated with the stuff, we'd have spotted it long ago.
How many people are directly exposed to it daily? Technicians and performers are probably it. Everyone else is very rare so it's possibly any side effects took a while for medical community to pick up on until everyone started vaping.
>At large, vapes are better than cigarettes.
Better yes, they are harm reduction over cigarettes. However, it's not "good" and should be as regulated as cigarettes are.
> There is one study looking at the potential to use PG as a carrier for an inhaled medicine (https://www.ncbi.nlm.nih.gov/pubmed/18158714) and another which mentions that PG or ethanol may be used as a cosolvent (https://www.ncbi.nlm.nih.gov/pubmed/12425745) in nebulizers, but no evidence presented of an asthma inhaler or nebulizer that is actually used today containing PG.
Even then, there's a huge difference between "being on stage with a fog machine", and 3-4 puffs a day of a smaller amount of a nebulizer, than chronic hundreds of puffs a day with vapes.
> Meta (and social media more broadly) are the only case where we have (in my opinion) substantiated allegations of a company being aware of a large, negative impact on society (mental wellness, of teens no less), and still prioritizing growth and profit
Them doing nothing about hate speech that fanned the flames for a full blown genocide is pretty terrible too. They knew the risks, were warned, yet still didn't do anything. It would be unfair to say the Rohingya genocide is the fault of Meta, but they definitely contributed way too much.
> Meta are the only case where we have substantiated allegations of a company being aware of a large, negative impact on society
Robinhood has entered the chat
Why would one specific industry be better? The toxic people will migrate to that industry and profit at the expense of society. It’s market efficiency at work.
I do think an industry is often shaped by the early leaders or group of people around them. Those people shape the dominant company in that space, and then go off to spread that culture in other companies that they start or join. And, competitors are often looking to the dominant company and trying to emulate that company.
> Early OpenAI set the tone of safe, open-source AI.
Early OpenAI told a bunch of lies that even (some of) their most-ardent fans are now seeing through. They didn't start off good and become the villain.
Gamifying day trading is just turning the retail market into gambling. Obvious objections will be that this has been possible for a long time now. But never did I know young men to casually play the market day to day like Wall Street Bets do now the way they would follow sports in the past.
Exploring unsophisticated investors. Trading on margin used to be for extremely experienced and educated people working for a large financial institution. The risk of margin trading is extreme with unlimited losses.
Also, tobacco companies and oil companies famously got into trouble from revelations that they were perfectly aware of their negative impacts. For the gambling and alcohol industry, it probably wouldn't even make the news if some internal report leaked that they were "aware" of their negative impact (as if anyone thought they would not be?)
Social media is way down on the list of companies aware of their negative impact. The negative impact arguably isn't even central to their business model, which it certainly is for the other industries mentioned.
The leaders and one of the announcers of Radio Télévision Libre des Mille Collines got 30 years to life sentences for their part in the Rwandan genocide.
If I remember correctly, other research has shown that it's not just the addictive piece. The social comparison piece is a big cause, especially for teenagers. This means Instagram, for example, which is highly visual and includes friends and friends-of-fiends, would have a worse effect than, say, Reddit.
Harakat (diacritics) are actually pretty rare in _any_ Arabic writing, unless it's the Quran (because meaning needs to be very exact/nuanced) or educational (for kids or other learners of the language). In fact, I think historically, not just the diacritics but even the dots (eg ب or ت) would be skipped, and people would guess the letter and pronunciation based on the context.
But you're right, with transliteration, it's much harder to guess because the sounds/combinations of letters are not typical, and the words are unfamiliar. So you just guess a bit and then you get corrected when you hear the sound (eg, on the song).
> Microsoft said that using Azure in this way violated its terms of service and it was “not in the business of facilitating the mass surveillance of civilians”. Under the terms of the Nimbus deal, Google and Amazon are prohibited from taking such action as it would “discriminate” against the Israeli government. Doing so would incur financial penalties for the companies, as well as legal action for breach of contract.
Insane. Obeying the law or ToS, apparently, is discriminatory when it comes to Israel.
U.S. law. It's pretty obvious that neither Amazon nor Google are good options for serious actors that are not the U.S. government. So if they want to make business outside the U.S., they need to dance around the fact that in the end they bow to the will of Washington.
It's not insane, at least based on the information in the article, which is entirely insinuation. Do we actually have access to the leaked documents and what specifically was being asked besides a "secret code" being used?
It would be suicide to sign the contract. It basically allows them to hack their platforms without any repercussions or ability to stop it. They would quickly claim expanded access is part of the contract.
This endless bowing down to Israel is and always will be ridiculous. When a country can do whatever they like unchallenged, no matter how wrong, or how illegal, we have failed as a society.
Because they overplayed their hand and they know it, so the only thing left to do is go all in and hope the walls they built hold long enough for this to be a fait accompli.
while the comment you reply to is borderline insane,
you're taking from a very privileged position in terms of media consumption. the media that criticizes the genocide and the blackflag on oct 7th is very niche and you seem to consume it exclusively. the message is very different within mass media.
I doubt the Guardian has any reason to lie about the documents they have seen. Based on the interactions regarding their war crimes, are you arguing Israel have not basically declared themselves above the law in many ways?
it doesn't matter what it does, why it's there, or how often it's used because: 1) skirts the law, 2) infringes on the laws of other countries, and finally 3) it's just so dodgy you have to be asking yourself wtf is going on.
How can an independent state "infringe on the laws of other countries"? If you think Israel is somehow bound by foreign states' laws, should it also be enforcing the Great Firewall, for example?
And how is it dodgy to want to know who spies on your data?
It is Israel's method introduced so that when Google and Microsoft who are legally required to pass over stored data based on where their servers are based, to find out who asked for it. I assume in the goal of trying to influence who asked for it.
1. You can access those models via three APIs: the Gemini API (which it turns out is only for prototyping and returned errors 30% of the time), the Vertex API (much more stable but lacking in some functionality), and the TTS API (which performed very poorly despite offering the same models). They also have separate keys (at least, Gemini vs Vertex).
2. Each of those APIs supports different parameters (things like language, whether you can pass a style prompt separate from the words you want spoken, etc). None of them offered the full combination we wanted.
3. To learn this, you have to spend a couple hours reading API docs, or alternatively, just have Claude Code read the docs then try all different combinations and figure out what works and what doesn't (with the added risk that it might hallucinate something).
reply