"Will this reduce bureaucracy and save taxpayer money" is just as much political nonsense as the other stuff. Taxpayer money unspent is not an unalloyed good. Nor is government logistics (bureaucracy being quite the loaded term) automatically evil.
Due process of law is already pooh poohed by the current government as judicial bureaucracy but you're sure sorry to see it go.
$1 trln+ of dollars on defence is not "nonsense". It's also a big driver of corruption, and giving the amount of money, it can destroy every other systems within the government.
Having AI in the mix could potentially fix the problem(partially).
AI is a big driver of literal direct physical corruption. Language and knowledge is forever tainted because of an outpouring of AI generated spam. Evaluating resumes is more difficult now because you can't tell real impacts versus fabricated hallucinations. Open source projects are overwhelmed with AI generated PRs...
Any corruption is emboldened by AI, it's a catalyst of the problem, doesn't seem anywhere close to potentially being a fix
An unbiased AI with access to all the military contracts would likely result in a lot of upset contractors who are used to scamming the government for as much as they can get away with. Grok by Musk (sounds like an awuful perfume!) is biased by design, not to mention the conflicts of interest.
That's a pretty wild false dichotomy; you think that the only possible option for reducing corruption in the government in the defense industry is paying billionaires for their AI products?
> Having AI in the mix could potentially fix the problem(partially).
Any examples?
As far as I understand, claims in the current AI cycle are wildly exaggerated, and sometimes companies rely on sort of circular deals to make revenue appear higher than it actually is, e.g. OpenAI and Microsoft or Nvidia. Wouldn't that mean that AI companies are primed to oversell and underdeliver, effectively making the problem even worse?
Has the author (not OP) written anything on this topic themselves? This is a blunt comment, because I am fed up with being asked to read LLM content that the prompter thinks is novel and worthwhile because they don't know better.
I can forgive (even root for) someone who puts in the effort themselves to understand a problem and write about it, even if they fall short or miss. They have skin in the game. I have little patience for someone who doesn't understand the disproportionate burden generated content places on the READER.
I can certainly tell they've put the model through the ringer to be terse and use simple language, etc. But I am struggling to separate the human ideas from the vibed ones, and the tone of the whole thing is the usual LLM elevator pitch with "hushed reverence" * "movie trailer cadence".
But "spawn/fork" is just a different way of labeling the fairly-well-understood tactic (I won't call it a strategy) of just how much context to provide sub-agents. Claude Code already "spawns" everytime it does an explore. It can do this concurrently, too.
Beyond that, they seem to express wonder at how well models can use tools:
> In the example above, the agent chose spawn for the independent research tasks and fork for the analysis that needs everything. It made this choice on its own — the model understands the distinction intuitively.
Emphasis mine. They (or the model whose output they blindly published) are anthropomorphizing software that is already designed to work this way. They gave it "fork" and "spawn" tools. Are they claiming they didn't describe exactly how they were supposed to be used in the tool spec?
The criticism about the labeling is valid and I oversold. For clarity, this is what the agent sees:
`spawn`: "Create a spawned child node under your node."
`fork`: "Create a forked child node (inherits parent context) under your
node."
The novelty is less about the distinction between the two, it's the tree generation. I would have served you better, if I just left out the parts that aren't critical to the novelty. Thank you for taking the time to comment.
In all honesty, "would have written it myself but I was too eager to get it out the door" doesn't really make sense to me. You're acknowledging that you took a shortcut to get it out the door (blog post as tech debt is a new one!) - does that mean you'd like to write something up yourself eventually?
I hope so, and would like to read it. In particular, since this is presented as research, I'd be very interested to read about your experimental observations that show the risks/costs/edge cases/limits of this pattern. As it stands, there's no indication of self-critique or really any process of systematic evaluation of the solution.
I enjoyed this! Thank you for taking the time to write it. I like it because I identify with your experiences way more than I do with the standard AI braggy boilerplate. Cheers.
And I deserve and accept the gentle snark at the beginning. I will be sure to let you know when I post something, I'll take your notes :)
on Wednesday, I had a dream about agents. Thursday evening, I talked to Claude about trees. That same night, I pushed out the post. There wasn't much rigor involved but yes I will explore more and report back to you!
I have not seen many of these hypothetical "intrinsically interesting" Show HN posts generated by unskilled users.
The single one I can think of is someone who (I quote) "accidentally created the fastest CSV parser ever using SIMD". This person had no interest in researching prior art themselves, and thus incorrectly claimed credit for "coming up" with this approach - and they didn't even do that.
It's not only the prose that's the problem if submitters are determined not to think.
At the code level it's still rehashing the same ideas over and over again. I wrote lots of things from software 3d on a weird system to jit to websites to telephony software to compilers to firmware for hardware to cloud orchestration and many other things and none of this was novel - someone wrote every single pattern from them before even if nobody put them together the same way. Putting known pieces together is not novel. And as a proportion, almost all software produced is just business apps of various types, with absolutely nothing novel in them.
Also from actual researchers, I know just one person who did something actually novel and it was with queuing.
> At the code level it's still rehashing the same ideas over and over again.
I agree that rehashing the same ideas over and over again is sufficient - for some strange, complacent definition of the word. It's not the only way to think about the discipline, and thank goodness enough smart people realize that.
> Also from actual researchers, I know just one person who did something actually novel and it was with queuing.
Think how many people have to be trying at any given time for it to happen at all.
Originally called Retriever, based on the domain. Trademark issues?
So it's a bunch of tools that Gemini can call, but the tools involve low-level interactions with the page structure in the end-user's browser.
What is the moat? What is an "agent" when you take away the powerful LLM?
Rover lives inside your website
Rover does not just live "inside" my website, because you are using Gemini 3 Flash to do all the heavy lifting.
Who is the audience here? It sounds like you're addressing people who don't know how the technology works, but the cutesy concept is borderline misleading.
Also, can you back up this claim with a human-written response? (emphasis mine)
When rtrvr.ai interacts with a webpage, there is zero automation fingerprint:
No navigator.webdriver flag
No CDP-specific JavaScript objects
*No detectable automation patterns in network requests*
*Identical timing characteristics to human interaction*
So our core technical moat is building up an agentic harness that can represent and take actions on any webpage without any screenshots. With this approach we even beat custom trained models like OpenAI Operator and Anthropic CUA:
https://www.rtrvr.ai/blog/web-bench-results
Everyone else in the space just takes a screenshot and asks a model what coordinates to click, our core thesis is that LLMs understand semantic representations fundamentally better than vision. But with this DOM approach there is a long tail of HTML/DOM edge cases to cover that we have built out for with the 20k+ users bringing these edge cases.
Soon you will be able to record demonstration tasks via our partner Chrome Extension as well as setup knowledge bases scraped by our Cloud browsers to provide additional context to the agent. So there is a platform moat as well.
The audience is website owners who want to increase visitor engagement and conversion via a conversational interface for users.
This is more for our cloud browser platform where we launch cloud browsers for vibe scraping controlled via a custom extension instead of CDP. You can try it out at rtrvr.ai/cloud, where we can get back some strong antibot detection sites like google.com
I'm very puzzled at the positive comments it's getting. It's insanely, discourteously long for the number of distinct ideas in it. There's banal LLM-ish smirking quips, very little personality and a lot of repetition.
There's certainly no personal anecdotes of difficult problems which would go a long way to show the author actually knows their stuff.
Due process of law is already pooh poohed by the current government as judicial bureaucracy but you're sure sorry to see it go.
reply