Does it seem, I'm not sure, ironic maybe? That the main example here is "An app that writes blog posts" - "Researches a topic and writes a blog post about it" - that the company who helped champion the network effects of the internet and surface truly useful search results is now helping to destroy that very same thing they built their entire business on?
Different teams. This AI team at Google cares not for the health of the web. They barely remember that Google has a search engine, except of course for the groudbreaking AI search results which are responsible for many large numbers for use in annual reviews. This team's goal is to sell their AI solution, and if that means demonstrating its ability to generate tools that create crap content that harms the search engine results, well, I'm sure another AI solution can probably combat that later.
Google Search has like 10 years of history of doing its best not to get you to click on a search result, but to answer your question directly or at the very least keep you on their platform while screwing over website owners.
The first two iteration of this were AMP and Instant Answers, the third one is AI Overview. AI Overview should not be seen in isolation, but as a part of the pattern. If it weren't for it, Google would double down on some other method of reaching the same goal.
This one will end up the same way the other two did: there's gonna be a vocal minority that's gonna consider it unfair and a web killer, the vast majority of users won't have an opinion, Google will not care, "the web" will play along, those early adopters are temporarily gonna have an advantage in this "new age" and some will die in the process, but the vast majority is gonna continue on as if nothing happened.
It's also not gonna be the final iteration of this process because shiny new things sound better to investors than marginal improvements, so X years from now AI Overview is gonna be seen as something "old-fashioned", Google Search will pivot once again, and the rest of the web will follow to keep Google happy.
Which is saddening as the first thing I think when I see this overview is "How do I verify this statement is correct" and paradoxically it sometimes just slows me down.
> the vast majority of users won't have an opinion
They're here, they don't care how they get from point A to point B, the tech used to achieve that result is completely irrelevant to them. AI? Great. Not AI such as the Instant Answers era? Also great. Average Joe does not spend his time thinking about the economics of the web.
But you shouldn't confuse them finding "AI" useful now with them being attached to it long term. It's a hip new tool now, but the novelty will fade and Google will have to re-invent themselves all over again. If anything, they kinda screwed themselves over by calling this "AI". AI is supposed to be something within reach, but always some years away. By wasting that term for the current era, it's gonna suck so hard to think of a new marketing term that's gonna be seen as an improvement in comparison to the term "AI".
Yeah hate to say it, because I am an AI hater, but I love the AI results in Google and Kagi. I barely click results anymore for basic questions unless it's something important enough for me to need verification to ensure the AI-gen answer wasn't a hallucination. It's been so nice not having to pick through the cesspool that is StackOverflow to find answers to quick cli questions, or wade through SEO-generated, Amazon-affiliate link garbage for more general questions.
This is what the vast majority of users will do of course. The issue people have with it is that it breaks the "social contract" of the web, which is that part of the advertising income goes to the site that provided the information the answer is based on. That, by destroying that income, "overviews" (now including AI overviews, but that's not where it started) are destroying first publishers and I'm sure it'll go all the way through until Youtube is entirely destroyed as well.
Of course, it does not destroy Google's income ... and it destroys the promise Google made long ago, which is to never keep users on Google platforms.
Oh and to add insult to injury, want to bet Opal will force app developers into what you probably never even imaged would happen on the web? Pay-per-view. Not for a video. For a website/app.
> This AI team at Google cares not for the health of the web. They barely remember that Google has a search engine
Neithet do any teams at Google, including the search teams. Google is an ad company with 80% of its revenue coming from ads. They couldn't give two shits about the health of the web.
That contradicts the long-standing claim that Google does care about as many people using as good a WWW as possible. A ghost town has plenty of ad space, but not much ad revenue.
Arguably the ad business is to blame. It created a perverse incentive. They maximized pay-to-play. The losers were authors that previously published on a passion budget (and would/could never pay for ads). AI is just the last nail in the coffin.
I dont think so. They also triggered SEO race where businesses pump out same bland blogposts to optimize ranking. Content made by humans for those companies was the only viable way at that time, and now new synthetic method emerges - whatever generates revenue will win. AI reels and tiktoks get views, so why bother with human generated content after the training on models have been done? Sad but true.
That's a good point, we had sort of the precursor to this already and yea likely driven by google themselves. It seems that every time incentives are aligned purely for profit we end up with situations like this where they inevitably run a good thing straight into the ground.
Funny, I remember the Internet from over a decade (or two) ago and it was a mess of full-screen ads, seizure-inducing animations, infinite popups, etc. that Google helped eliminate.
Google didn’t eliminate malicious ads - adblockers and easylists did that.
I hate to break it to you bud, but the full-page ads never went away - they just look like content now. You know why you need to scroll for 9 pages to see the ingredients to a recipe?
Google created the pageview driven business model that incentivized the internet to be filled with hostile UX, low-quality lists of paged content, affiliate spam, etc.
> that the company who helped champion the network effects of the internet and surface truly useful search results...
The amount of data on the web crossed the threshold of organic discoverability some time before the AI boom started. AI makes it go from really bad to really, really bad (99% to 99.99%). As far as I am concerned it doesn't change anything.
The same mechanisms to find good content would work today as well - following humans and networks.
If they store both the generated content and the eventual indexed location, they could now filter search results more comprehensively based on content hashes.
They can't even monopolize the ai effort in their own org. There's a dozen google ai products that all compete with each other (ai studio, firebase studio, opal, Gemini etc).
They have all the potential but have lost direction years ago.
I think they will stay relevant but not dominant, much like the case of Google meet.
I’ve never thought of Google as having some clear direction short of having dominated with the pagerank search.
It reminds me of people who invented a thing that is ubiquitous and therefore just sits there printing money all their life, but basically every single other thing they’ve done since then has been at abysmal returns on capital (ie, where they would have been better off just not doing anything and putting it in the money market… their contribution was extremely negative), while they delude themselves into what wonderful success they’ve been because the one-shot invention keeps printing money.
One exception might be YouTube, but if I recall the story correctly, Google was going to shut it down too when one guy (I don’t recall now) made a proposal and single-handedly turned it into the powerhouse it now is. But that’s just dumb luck on Google’s part. That guy could have also just as well have been somewhere else at that pivotal point and Google would have killed off YouTube like it has killed off so many other projects because it lacks the actual competence to follow through with things… everything always remains with a sense of being half-baked.
That reminds me of a thing I stumbled upon a while ago now, that technically speaking several of Google’s core offerings are still designated as Beta projects behind the scenes. I’m not sure why exactly, but even if it’s just an oversight of background clutter, that seems symbolic of Google; the disheveled, disorganized, cluttered personality that has not resolved loose ends smothered in decades of other loose ends.
I doubt that, I hear on the internet that Gemini pro is great but every time I have used it has been beyond disappointing. I’m starting to believe that the Gemini pro is great is some paid PR push and not based on reality. The Gemma models are also probably the least useful/interesting local models I’ve used.
What are you using them for? Gemini (the app, not just the Google search overview) has replaced ChatGPT entirely for me these days, not the least of which is because I find Gemini simply be able to handle web searches better (after all, that is what Google is known for). Add to that, it can integrate well with other Google products like YouTube or Maps where it can make me a nice map if I ask it what the best pizza places are in a certain area. I don't even need to use pro mode, just fast mode, because it's free.
Claude is still used but only in IDEs for coding, I don't ask it general questions anymore.
I use Gemma as a developer for basic on-device LLM tasks such as structured JSON output.
That's true but to be honest I didn't really use those features anyway, my chats are just one long stream of replies and responses. If I need to switch to a new topic I make a new chat.
I used Gemini Pro and it was unable to comply with the simplest instructions (for image diffusion). Asking it to change the scene slightly by adding or removing object or shifting perspective yielded almost the same result, only with some changes I did not ask for.
The image quality was great, but when I ask a woodworker for a table and get a perfectly crafted chair of the highest quality, I'm still unsatisfied.
I cancelled my subscription after two days trying to get Gemini to follow my instructions.
1. "it's not available in your country yet" -- although I am currently in (and connected to) Czech republic train, my account is based on Luxembourg. Not sure which one takes precedence, but sad to see...
2. Join Discord -- Unexpected to see this from a Google product. More interestingly, do they really have people/staff there? Or is it just bunch of AI Agents running the discord server, not sure. (Haven't joined either)
It's been available in Vietnam for around 2 weeks now. So maybe it's a privacy thing in the EU?
Honestly though you're not missing much, it doesn't feel revolutionary.
> Join Discord -- Unexpected to see this from a Google product
I did a trial of Gemini Enterprise last month - a product absolutely not ready to be released - and they use Slack which also surprised me. So maybe internal teams at Google are allowed to choose their own messaging platform?
Slack is expected as Kubernetes/Cloud-Native/CNCF already used Slack. They also had extensive set of bots and integrations, including GitHub workflows too.
It showed up as "a new way to build Gems" on my Gemini account a couple of weeks ago. I tried it a couple of times but the basic idea I had of creating a top level agent with sub-agents depending on the query (supervisor pattern I think?) doesn't seem to work. There's no branching, it seemed to just run app possible paths at the same time which takes ages and wastes tokens. Maybe I was doing something wrong, or maybe I misunderstood the purpose of this. But each thing I tried could be handled better by a normal gem with custom prompt.
It might be useful for a specific kind is multi step non branching problem, but I didn't have any problems like that to test it with.
Reminds me of a quote from a few years back: "We are entering an era where we use AI to write blog posts from a few keywords for people who use AI to summarize a blog post into a few keywords".
Not dismissing your first argument but we had good success hiring talent through Discord servers. If you want to reach the same audience that attends hackathons and such it's in my experience the best platform.
Do you mean Discord is a kind of community for young hackers? I’m curious about Discord because I’ve been asked about a chat app for teenagers who don’t have mobile phones.
They've used discord for a while for things like GSoC but in general it's becoming more widespread seeing that even Copilot has a Microsoft owned discord server, Valdi from Snapchat as well.
I didn't get far with this because it wants access to my entire Google Drive, which I declined. Credit to Google for even offering the chance to say "no", I suppose.
So you trust Google with the data in your google drive, but you don't trust Google (Opal Team) with the data in your drive?
Yes.
More specifically, I trust Google not to use my files to train its AI if I haven't given permission, but I don't trust Google not to use Opal as a way to get me to give them permission without realising.
How do I even know this is a real Google product? Okay I’ll trust the domain (hopefully it’s not googIe.com). How do I know it has no vulnerabilities? Google is a massive company. There’s a big difference between trusting an established team vs. whoever this team is.
Expecting permissions to my entire Google Drive is ridiculous. Yes, I tried not granting that permission (and only granting permission to an app-specific path) and it specifically told me I have to grant full permission . I closed the tab.
>How do I even know this is a real Google product? Okay I’ll trust the domain
it is not even the expected opal.google.com it is opal.google, you need to be 100% sure beforehand that google has the sole rights to the .google tld (which an average person wouldn't know)
The behavior also seems sketchy with it asking for permission but then rejecting any usage if all of the permissions are not approved (why even ask then, you are google)
after re-finding the link through a confirmed subdomain.google.com site I tried to sign in and got this error
```
An unexpected signin error occured.Error checking geo access
```
I didn’t even notice the TLD, wow that makes it even worse. Funny thing is I don’t actually have anything in my Gdrive. It’s just the principle that irks me.
I think the concern is that this might somehow enable a privacy policy they weren't aware of that permits training over the entire Drive. However, I think the primary reason for this is that these products generally would like to store data on the user's Google Drive but Google Drive doesn't have super granular permission structure to be able to set up a partitioned directory for the app alone. I actually think that might be a good thing to work on next?
That was my concern too. However, the provided links to both the ToS and privacy policy were the standard Google ones (https://policies.google.com/terms), so it seems not to be giving Opal special privileges to read/train on Drive data.
Do any of the example apps work for anyone? I tap “try now” on one, and it just opens a page with its logo/name/description. There’s a sidebar menu that just has its name, and a restart app button that does nothing. I can’t see how to make the app do anything.
I had to switch from Duck Duck Go browser to Safari, then the Book Recommendation app worked for me. For what its worth, the recommendations were good, evidenced by my having already read half the recommended books. The vibe promots to create this app were incredibly simple.
I wonder how android developers feel about this. You put years of your life into supporting a companies product, and then they actively try and remove the need for your job to exist.
that is because they are all using non deterministic approaches, aka expecting that a single detailed prompt with 10000 words is going to generate a stable application. Because prompts dont have replay value, you have to split it into one microtask per agent and validate the output with deterministic fallback as and when required.
See them in the Google Graveyard most likely. This is one of those G experiments for which I wouldn't invest 5 minutes of my time building something, knowing it will surely be "sunsetted" in a short time.
They had an invite only one in the bard days before it was rebranded to gemini. You didn't just need to get the invite link but actually link your discord to your google so I didn't bother.
Just goes to show that google's attempts at chat have been a big flop and even though google chat exists they don't use it.
These companies won’t rest until they’ve figured out how to get rid of all their actual employees and have replaced their business strategy with “just click YES on whatever the slop machine cranks out this quarter”
Does it seem, I'm not sure, ironic maybe? That the main example here is "An app that writes blog posts" - "Researches a topic and writes a blog post about it" - that the company who helped champion the network effects of the internet and surface truly useful search results is now helping to destroy that very same thing they built their entire business on?
Different teams. This AI team at Google cares not for the health of the web. They barely remember that Google has a search engine, except of course for the groudbreaking AI search results which are responsible for many large numbers for use in annual reviews. This team's goal is to sell their AI solution, and if that means demonstrating its ability to generate tools that create crap content that harms the search engine results, well, I'm sure another AI solution can probably combat that later.
Google Search has like 10 years of history of doing its best not to get you to click on a search result, but to answer your question directly or at the very least keep you on their platform while screwing over website owners.
The first two iteration of this were AMP and Instant Answers, the third one is AI Overview. AI Overview should not be seen in isolation, but as a part of the pattern. If it weren't for it, Google would double down on some other method of reaching the same goal.
This one will end up the same way the other two did: there's gonna be a vocal minority that's gonna consider it unfair and a web killer, the vast majority of users won't have an opinion, Google will not care, "the web" will play along, those early adopters are temporarily gonna have an advantage in this "new age" and some will die in the process, but the vast majority is gonna continue on as if nothing happened.
It's also not gonna be the final iteration of this process because shiny new things sound better to investors than marginal improvements, so X years from now AI Overview is gonna be seen as something "old-fashioned", Google Search will pivot once again, and the rest of the web will follow to keep Google happy.
you know, there is a third group - a large number of users which find the AI overview useful
Which is saddening as the first thing I think when I see this overview is "How do I verify this statement is correct" and paradoxically it sometimes just slows me down.
> the vast majority of users won't have an opinion
They're here, they don't care how they get from point A to point B, the tech used to achieve that result is completely irrelevant to them. AI? Great. Not AI such as the Instant Answers era? Also great. Average Joe does not spend his time thinking about the economics of the web.
But you shouldn't confuse them finding "AI" useful now with them being attached to it long term. It's a hip new tool now, but the novelty will fade and Google will have to re-invent themselves all over again. If anything, they kinda screwed themselves over by calling this "AI". AI is supposed to be something within reach, but always some years away. By wasting that term for the current era, it's gonna suck so hard to think of a new marketing term that's gonna be seen as an improvement in comparison to the term "AI".
Yeah hate to say it, because I am an AI hater, but I love the AI results in Google and Kagi. I barely click results anymore for basic questions unless it's something important enough for me to need verification to ensure the AI-gen answer wasn't a hallucination. It's been so nice not having to pick through the cesspool that is StackOverflow to find answers to quick cli questions, or wade through SEO-generated, Amazon-affiliate link garbage for more general questions.
This is what the vast majority of users will do of course. The issue people have with it is that it breaks the "social contract" of the web, which is that part of the advertising income goes to the site that provided the information the answer is based on. That, by destroying that income, "overviews" (now including AI overviews, but that's not where it started) are destroying first publishers and I'm sure it'll go all the way through until Youtube is entirely destroyed as well.
Of course, it does not destroy Google's income ... and it destroys the promise Google made long ago, which is to never keep users on Google platforms.
Oh and to add insult to injury, want to bet Opal will force app developers into what you probably never even imaged would happen on the web? Pay-per-view. Not for a video. For a website/app.
Very large even. Especially considering the explosion of ai-crap in the search results
HN's hate for LLM based AI blinds it to this uncomfortable possibility
> This AI team at Google cares not for the health of the web. They barely remember that Google has a search engine
Neithet do any teams at Google, including the search teams. Google is an ad company with 80% of its revenue coming from ads. They couldn't give two shits about the health of the web.
The man who killed Google search was a minor hit when it was published: https://www.wheresyoured.at/the-men-who-killed-google/
That contradicts the long-standing claim that Google does care about as many people using as good a WWW as possible. A ghost town has plenty of ad space, but not much ad revenue.
Don't look at what a company claims. Look at what the company does.
By that logic no company cares about anything but profits.
You say that like it’s this big contradiction of some obvious truth, so go on, tell us about this company you know of that Really Cares.
How is this surprising? The absolute vast majority of companies care about nothing but profits.
As for Google, what I wrote was literally spelled out in their own emails uncovered in court proceedings.
Arguably the ad business is to blame. It created a perverse incentive. They maximized pay-to-play. The losers were authors that previously published on a passion budget (and would/could never pay for ads). AI is just the last nail in the coffin.
I dont think so. They also triggered SEO race where businesses pump out same bland blogposts to optimize ranking. Content made by humans for those companies was the only viable way at that time, and now new synthetic method emerges - whatever generates revenue will win. AI reels and tiktoks get views, so why bother with human generated content after the training on models have been done? Sad but true.
That's a good point, we had sort of the precursor to this already and yea likely driven by google themselves. It seems that every time incentives are aligned purely for profit we end up with situations like this where they inevitably run a good thing straight into the ground.
Google has already been single-handedly destroying the internet for over a decade by turning into an ad-ridden mess
Funny, I remember the Internet from over a decade (or two) ago and it was a mess of full-screen ads, seizure-inducing animations, infinite popups, etc. that Google helped eliminate.
Google didn’t eliminate malicious ads - adblockers and easylists did that.
I hate to break it to you bud, but the full-page ads never went away - they just look like content now. You know why you need to scroll for 9 pages to see the ingredients to a recipe?
Google created the pageview driven business model that incentivized the internet to be filled with hostile UX, low-quality lists of paged content, affiliate spam, etc.
We've had 25 years to regulate them or break them up. It's our own fault for under-regulation.
It is never too late to make user tracking illegal and destroying the nastiest part of the ad industry
Perhaps we should look to Europe for inspiration on how to govern tech companies.
> that the company who helped champion the network effects of the internet and surface truly useful search results...
The amount of data on the web crossed the threshold of organic discoverability some time before the AI boom started. AI makes it go from really bad to really, really bad (99% to 99.99%). As far as I am concerned it doesn't change anything.
The same mechanisms to find good content would work today as well - following humans and networks.
If they store both the generated content and the eventual indexed location, they could now filter search results more comprehensively based on content hashes.
Yeah, helping to produce AI-generated garbage.
"Opal is not available in your country yet"
Thanks for having me click through 5 screens including giving access to Google Drive to tell me that in the end.
Isn't this the same thing they showcased as part of the Gemini 3 Flash release some time ago?
On https://blog.google/products/gemini/gemini-3-flash/ , the paragraph that starts with:
"Or you can quickly build fun, useful apps from scratch using your voice without prior coding knowledge."
Yeah me too. 6 months after original launch! What a crappy product experience, who's managing this rubbish?
I think Google is going to do with consumer AI as they have done with search engine. Full monopolization. They own the lumberyard and the forest.
They'll just see whats popular and then clone, launch and instantly own verticals.
It's over for the little SaaS guys.
https://www.youtube.com/watch?v=BxFQYw_MmAA
They can't even monopolize the ai effort in their own org. There's a dozen google ai products that all compete with each other (ai studio, firebase studio, opal, Gemini etc).
They have all the potential but have lost direction years ago.
I think they will stay relevant but not dominant, much like the case of Google meet.
> all compete with each other
It's common business practice to set up internal innovation competitions, and blend the best.
You call all these products launches from google "internal innovation competitions"?
And even if they were, which they aren't, are you sure it's a "common business practice"? How many companies can afford that.
> lost direction years ago.
I’ve never thought of Google as having some clear direction short of having dominated with the pagerank search.
It reminds me of people who invented a thing that is ubiquitous and therefore just sits there printing money all their life, but basically every single other thing they’ve done since then has been at abysmal returns on capital (ie, where they would have been better off just not doing anything and putting it in the money market… their contribution was extremely negative), while they delude themselves into what wonderful success they’ve been because the one-shot invention keeps printing money.
One exception might be YouTube, but if I recall the story correctly, Google was going to shut it down too when one guy (I don’t recall now) made a proposal and single-handedly turned it into the powerhouse it now is. But that’s just dumb luck on Google’s part. That guy could have also just as well have been somewhere else at that pivotal point and Google would have killed off YouTube like it has killed off so many other projects because it lacks the actual competence to follow through with things… everything always remains with a sense of being half-baked.
That reminds me of a thing I stumbled upon a while ago now, that technically speaking several of Google’s core offerings are still designated as Beta projects behind the scenes. I’m not sure why exactly, but even if it’s just an oversight of background clutter, that seems symbolic of Google; the disheveled, disorganized, cluttered personality that has not resolved loose ends smothered in decades of other loose ends.
I doubt that, I hear on the internet that Gemini pro is great but every time I have used it has been beyond disappointing. I’m starting to believe that the Gemini pro is great is some paid PR push and not based on reality. The Gemma models are also probably the least useful/interesting local models I’ve used.
What are you using them for? Gemini (the app, not just the Google search overview) has replaced ChatGPT entirely for me these days, not the least of which is because I find Gemini simply be able to handle web searches better (after all, that is what Google is known for). Add to that, it can integrate well with other Google products like YouTube or Maps where it can make me a nice map if I ask it what the best pizza places are in a certain area. I don't even need to use pro mode, just fast mode, because it's free.
Claude is still used but only in IDEs for coding, I don't ask it general questions anymore.
I use Gemma as a developer for basic on-device LLM tasks such as structured JSON output.
Gemini just has many basic things missing like the ability to edit a message more than one message in the past and see branches of that conversation.
That's true but to be honest I didn't really use those features anyway, my chats are just one long stream of replies and responses. If I need to switch to a new topic I make a new chat.
The local Gemma models are pretty good for tasks involving multilingual inputs (translation, summarization, etc.). They have their niche.
I use Gemini from within AI Studio [0]. Not sure in what way you find Gemini disappointed, but I have get success with it through AI studio.
[0] https://aistudio.google.com
I used Gemini Pro and it was unable to comply with the simplest instructions (for image diffusion). Asking it to change the scene slightly by adding or removing object or shifting perspective yielded almost the same result, only with some changes I did not ask for.
The image quality was great, but when I ask a woodworker for a table and get a perfectly crafted chair of the highest quality, I'm still unsatisfied.
I cancelled my subscription after two days trying to get Gemini to follow my instructions.
Discussion when launched 5 months ago: https://news.ycombinator.com/item?id=44681786
The unresponsive search bar centred on the viewport is all I need to know about this product.
Google may deliver us the AI future Altman promised. The PM who thinks animated PNGs pass for anything real is not on that path.
I clicked the search bar thinking thats how i can try it.
Turns out its just an animation and there’s a button underneath to actually try it.
2 cents from me
1. "it's not available in your country yet" -- although I am currently in (and connected to) Czech republic train, my account is based on Luxembourg. Not sure which one takes precedence, but sad to see...
2. Join Discord -- Unexpected to see this from a Google product. More interestingly, do they really have people/staff there? Or is it just bunch of AI Agents running the discord server, not sure. (Haven't joined either)
The list of countries where its available is here: https://developers.google.com/opal/faq
Not a single EU country, so, yes.
It's been available in Vietnam for around 2 weeks now. So maybe it's a privacy thing in the EU?
Honestly though you're not missing much, it doesn't feel revolutionary.
> Join Discord -- Unexpected to see this from a Google product
I did a trial of Gemini Enterprise last month - a product absolutely not ready to be released - and they use Slack which also surprised me. So maybe internal teams at Google are allowed to choose their own messaging platform?
Slack is expected as Kubernetes/Cloud-Native/CNCF already used Slack. They also had extensive set of bots and integrations, including GitHub workflows too.
TIL, Google Labs use discord :)
Just short of 300k members on discord (almost 17k online) at the time of writing this. It's their Google Labs discord.
Maybe it's just me, but it doesn't inspire anything in me.
This idea of make-things-quick-without-any-real-skills seems fundamentally contrary to achieving lasting quality...
This line of thought does not apply to the tortilla, the roti and the naan.
YMMV
Anyone have comments actually about Opal? Curious if anyone tried it.
It showed up as "a new way to build Gems" on my Gemini account a couple of weeks ago. I tried it a couple of times but the basic idea I had of creating a top level agent with sub-agents depending on the query (supervisor pattern I think?) doesn't seem to work. There's no branching, it seemed to just run app possible paths at the same time which takes ages and wastes tokens. Maybe I was doing something wrong, or maybe I misunderstood the purpose of this. But each thing I tried could be handled better by a normal gem with custom prompt.
It might be useful for a specific kind is multi step non branching problem, but I didn't have any problems like that to test it with.
Yeah the comments section is really disappointing here. Sadly, I can’t try it because it’s blocked in UK. I’ll try again over VPN later.
"An app that writes blog posts" ... how about an app that reads blog posts? ^^
Reminds me of a quote from a few years back: "We are entering an era where we use AI to write blog posts from a few keywords for people who use AI to summarize a blog post into a few keywords".
“Join our Discord…”
That is a bit unexpected to see. Start Up vibes, haha.
Looks like another insane leap forward this.
In my estimation, that’s memetics and signaling to an in-group about who they want their audience to think their audience is.
That’s why I did it. Oh, and it’s nice to be able to chat with potential users. By why not Slack or Teams?
Not dismissing your first argument but we had good success hiring talent through Discord servers. If you want to reach the same audience that attends hackathons and such it's in my experience the best platform.
Do you mean Discord is a kind of community for young hackers? I’m curious about Discord because I’ve been asked about a chat app for teenagers who don’t have mobile phones.
They've used discord for a while for things like GSoC but in general it's becoming more widespread seeing that even Copilot has a Microsoft owned discord server, Valdi from Snapchat as well.
I didn't get far with this because it wants access to my entire Google Drive, which I declined. Credit to Google for even offering the chance to say "no", I suppose.
Why does it need that?
Opal isn’t just peeking at Drive, it’s using it as its backend. Outputs get saved as Drive files so they persist, can be shared, and open in Docs, etc
The gotcha is the permission scope can be pretty broad (read/write or metadata across Drive), so it’s worth checking what you actually granted.
So you trust Google with the data in your google drive, but you don't trust Google (Opal Team) with the data in your drive?
They need a place to store data, Google Drive is a place for that. Have you used NotebookLM or such which do the same sort of thing?
So you trust Google with the data in your google drive, but you don't trust Google (Opal Team) with the data in your drive?
Yes.
More specifically, I trust Google not to use my files to train its AI if I haven't given permission, but I don't trust Google not to use Opal as a way to get me to give them permission without realising.
How do I even know this is a real Google product? Okay I’ll trust the domain (hopefully it’s not googIe.com). How do I know it has no vulnerabilities? Google is a massive company. There’s a big difference between trusting an established team vs. whoever this team is.
Expecting permissions to my entire Google Drive is ridiculous. Yes, I tried not granting that permission (and only granting permission to an app-specific path) and it specifically told me I have to grant full permission . I closed the tab.
>How do I even know this is a real Google product? Okay I’ll trust the domain
it is not even the expected opal.google.com it is opal.google, you need to be 100% sure beforehand that google has the sole rights to the .google tld (which an average person wouldn't know)
The behavior also seems sketchy with it asking for permission but then rejecting any usage if all of the permissions are not approved (why even ask then, you are google)
after re-finding the link through a confirmed subdomain.google.com site I tried to sign in and got this error
``` An unexpected signin error occured.Error checking geo access ```
so I gave up
I didn’t even notice the TLD, wow that makes it even worse. Funny thing is I don’t actually have anything in my Gdrive. It’s just the principle that irks me.
Select what Opal can access
- "See and download all your Google Drive files."
I tried de-selecting that, and it told me to re-login and enable that setting.
The Google drive is also hosted by Google on Google's servers. They already have access to everything in there.
I think the concern is that this might somehow enable a privacy policy they weren't aware of that permits training over the entire Drive. However, I think the primary reason for this is that these products generally would like to store data on the user's Google Drive but Google Drive doesn't have super granular permission structure to be able to set up a partitioned directory for the app alone. I actually think that might be a good thing to work on next?
That was my concern too. However, the provided links to both the ToS and privacy policy were the standard Google ones (https://policies.google.com/terms), so it seems not to be giving Opal special privileges to read/train on Drive data.
It wants to add it to the training set ... (guess)
Opal is Optimizely AI name, but I guess Google's being Google, they are the ones that need to now change the product name.
https://www.optimizely.com/ai/
Here's the OSS repo for Google Opal: https://github.com/breadboard-ai/breadboard
Do any of the example apps work for anyone? I tap “try now” on one, and it just opens a page with its logo/name/description. There’s a sidebar menu that just has its name, and a restart app button that does nothing. I can’t see how to make the app do anything.
I had to switch from Duck Duck Go browser to Safari, then the Book Recommendation app worked for me. For what its worth, the recommendations were good, evidenced by my having already read half the recommended books. The vibe promots to create this app were incredibly simple.
I wonder how android developers feel about this. You put years of your life into supporting a companies product, and then they actively try and remove the need for your job to exist.
if your app can be replaced by a few vibe prompts I doubt it was successful in the first place
Not available in Germany and I assume the rest of Europe
"not available in your country" (I'm in Belgium)
Funny, everyone is trying to do the same thing. IMHO no one nailed it yet.
that is because they are all using non deterministic approaches, aka expecting that a single detailed prompt with 10000 words is going to generate a stable application. Because prompts dont have replay value, you have to split it into one microtask per agent and validate the output with deterministic fallback as and when required.
Is "an app to write blog posts" the best example they could find? What’s even the point, who will even read AI slop blog posts?
If they can’t even come up with a good, useful idea to showcase their product, I can’t believe it’s worth it to begin with.
> Not available in your country yet
See ya in 6 months !
See them in the Google Graveyard most likely. This is one of those G experiments for which I wouldn't invest 5 minutes of my time building something, knowing it will surely be "sunsetted" in a short time.
That will make 300
https://killedbygoogle.com/
> Join our Discord for support and sharing feedback
I’m surprised to see Google directing people to Discord, do they do that for other products?
They had an invite only one in the bard days before it was rebranded to gemini. You didn't just need to get the invite link but actually link your discord to your google so I didn't bother.
Just goes to show that google's attempts at chat have been a big flop and even though google chat exists they don't use it.
They did for the new copilot agent builder products, thats where I get most early invites.
They used to have a lot of IRC channels on freenode for Android.
I recall them taking about one for NotebookLM.
Jules, their cloud AI agent.
What is this AI mini app?
Do this make an actual production Flutter app or something?
It's basically an agent (Gem, as they call them) that you can run from your Gemini account.
First thing I thought was “oh, neat, another google product they will kill”.
Oh neat, another service they will hold hostage if they arbitrarily decide to lock my Google account.
"Platform roulette" was my first thought. Then I realised the chance of winning on this gamble is approximately nil.
Anyone taking bets on when this shows up on killedbygoogle.com ?
not available in your country...
I suppose for any in europe waking up like me that I can save you one click and some time.
These companies won’t rest until they’ve figured out how to get rid of all their actual employees and have replaced their business strategy with “just click YES on whatever the slop machine cranks out this quarter”
What is the usual life expectancy of a Google product?
Google usually kills projects. What's the point in using this?
"The horror, the horror..."
Problem I see with this being "codeless" is now Google owns everything. They can just hold your app hostage at any price point.
One more Google product destined to die!
Cute . I think we're all fine with anything that remotely threatens Vercel.