Hacker Newsnew | past | comments | ask | show | jobs | submit | snug's commentslogin

Really excited to try the gsuite replacements, which seems like the bigger announcement than the title suggest

Why? I only use gsuite for interop between Claude and ChatGPT and other services...so a moat like that is totally worthless to me

I still use Google Workspace for personal and work stuff, but for solo agent workflows, I think I have some use cases that would help me personally. I can let agents spin up working documents that I can review, comment on, and iterate over asynchronously.

For example, I’ve spent the last few weeks building an internal ticketing setup for my agents so they stop trying to fix every tangential issue they encounter. I wrote a skill where subagents bubble out-of-scope issues up to a parent orchestrator, which decides whether to log a ticket.

Currently, that setup mirrors between a local SQLite database and an Obsidian vault: agents write to SQLite, which syncs to Obsidian; when I edit notes in Obsidian, it syncs back. Native Claude docs mean I could ditch the Obsidian middleman for remote review—editing tickets from anywhere while keeping the core SQLite state local to where the agents execute.


That makes sense since you treat it as your central productivity manager, but for me all of Claude/Codex/etc are not productivity tools, just heterogeneous compute /for/ the productivity tool :P

All the gauges on the top right except the RPMs move


I'd wager a motorbike is much safer traveling at the same speed


Why do you say that?


I think this can be great as additional layer of security. Where you can have a non llm layer do some analysis with some static rules and then if something might seem phishy run it through the llm judge so that you don’t have to run every request through it, which would be very expensive.

Edit: actually looks like it has two policy engines embedded


And we don't think the judge can/will be gamed? Also... It's an LLM, it's going to add delay and additional token burn. One subjective black box protecting another subjective black box. I mean, what couldn't go wrong?


you can use a safety model trained on prompt injections with developer message priority.

user message becomes close to untrusted compared to dev prompt.

also post train it only outputs things like safe/unsafe so you are relatively deterministic on injection or no injection.

ie llama prompt guard, oss 120 safeguard.


Unfortunately it's not that simple. Self-policing AI systems will always be gamed. Just one [0] example of this among many.

[0] https://www.hiddenlayer.com/research/same-model-different-ha...


What happens when a prompt injection attack exploits the judge LLM and results in a higher level of attacker control than if it never existed?


How can it result in a higher level of control? I don't see why the "judge" should have access to anything except one tool that allows it to send an "accept" or "deny" command.


Pretty common to call something that isn't ready a preview


I've generally chosen these tools when I am creating a project. Though I generally use firebase hosting vs other front-end hosting. They have a much more generous free plan.

I'd suggest making some changes to how some of these things are categorized. You have database section with postgres at the top and then with supabase as number 2, but that's also a hosted postgres.

Overall, great job to the creators of this, I enjoyed seeing this analysis


I've also dove into all of those, and have mostly stuck to OpenSCAD now. I'm not amazing at it, but I've been able to get a few things done that I needed for 3D printing. What has really made much better at OpenSCAD is ClaudeCode or Antigravity in VSCode, with BSOL2 library. The documentation is just bad enough that it takes me forever to figure out on my own, but just good enough with lots of examples out there that an LLM can get mostly what I want with little fuss.


When I was in Italy, it took me far too long to realize on Google Maps that all the restaurants I was visiting started with "Il" and not the roman numeral two "II"


I've been able to get this same issue when I am laying in bed and lay on my side where my APP rub against the pillow. I've been able to reliably make the screeching sound by putting one airpod in my hand and tumbling it around or opening and closing my hands.

I still love my APP3, and still have my APP2, but have not made a change back to them. I think they feel so much better in my hear with the new tips, and sound way better.


You can downvote youtube videos without the extension, the extension just shows how many people downvoted a video since youtube took that visibility away a few years ago. It does help when there are shitty video recommendations and can ignore videos that have a bad ratio


Subtle correction: The extension shows how many users of the extension downvoted the video. It has no way of knowing what non-extension-users did.


The paid for feature seems to show the number of dislikes that have the extension installed, however what is shown is the estimated number of dislikes based on the actual number of likes youtube shows, so if the extension users have a like/dislike ratio of 50%, and there youtube shows 1000 likes, the extension estimates that there are also a 1000 users that disliked the video, even though there may only be 1 like and 1 dislike from the extension users, which would be very obvious that the estimate is very off.


IIRC, in the earliest days, the extension reported the dislike count that was still in the API but not surfaced in any clients. At some point after that, Google removed the count from the API entirely, requiring the current piecemeal approach.


That was true for about 3 weeks (ie. less than 2% of the extension's history)


Yes, but I believe it shows an estimate of the total downvotes based on the proportion of its own users who have liked vs disliked.


bad ratio can be caused by a lot of things


Still useful


I'm skeptical that it's of any use considering it's only going to be installed by people who specifically want to dislike. It seems more likely that the like/dislike ratio is going to be heavily skewed towards dislikes just by the stated purpose of the plugin and who it appeals to.


I don't think the extension is installed by people that wanna dislike, rather by people that wanna see the dislike count. I used to have it since it's very useful to see if something like a tutorial is a scam.

Also if I remember correctly, the extension was created before youtube completely removed the count, and it scraped the actual value from youtube to add it to the db while the count still was available


> I don't think the extension is installed by people that wanna dislike, rather by people that wanna see the dislike count.

Yes but the corollary is that the dislikes you see are an estimate extrapolated from dislikes from people who use the extension.

In other words if people want their dislike to be seen, they need the extension.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: