ashu1461

My point is about the emphasis of Codex in the title. That emphasis makes more sense when Codex is credited with finding something that would have been difficult or impractical to discover without substantial human effort.

What’s easier: waiting until your child is mature enough before giving them a smartphone, or trying to regulate social media companies and every addictive website?

In most of the cases, web apps are also much slower as well and the UX is also sub par.

Is it mainly about how the main/orchestrator agent communicates with its subagents ?

If desired the user can always see what the sub agent is doing in detail ?

Isn't it the same in case of claude as well ?

Someone used Codex to scrape the ICM website schedule and discovered that the winners list was simply hidden in the front-end code with a "hidden" tag

This is on the devs and feels like a very basic leak which could have exploited in the non LLM world as well.

They said in the AI community, a pelican riding a bicycle is a good test to measure effectiveness of the model, wondering if they were referring to you, or is it really a standard in the AI community ?

Also would be good to have a tool where users can select models and instantly see each model's generated pelicans. That will make it easy to compare the output of different models.

Right now even physical things have digital interlocking. Example you might purchase a watch but it requires you to purchase a subscription for it to function.

It is just the way most of the things are.

Looks like something which will be only helpful for bigger companies who have centralised auth. Good step nonetheless

Same we are also stuck with ES wish could migrate to clickhouse but not able to do so because of the legacy load.

Can you please expand more on the claude analysis part. What exactly you analysed and what outcome it helped with ?

Wondering if most of the AI agents use real time apis or transcription apis.. anyone had experience with building voice agents can comment ?

Isn't the article over emphasising a little bit on leakage of internal urls ?

Internal hostnames leaking is real, but in practice it’s just one tiny slice of a much larger problem: names and metadata leak everywhere - logs, traces, code, monitoring tools etc etc.

And not just any FTEs, probably few senior / staff level engineers who would cost a lot more.

Reminds me of the year 2038 problem, which will be more scary than this.

So if you are storing epoch time (no of milli seconds after 1970) in integer variables, your code will most probably break, which I feel is going to happen in a lot of scenarios.

Meditation used to need a cave or a mountain, now it just needs a cafe and no phone.

Google also used to have a go app which they deprecated later on, while now i think about it what is the use of having a go app, if the websites which are shown in search results are not optimised for slower networks.

I had the same doubt. With CLIs you can make your own custom shortcuts, LLMs can use it to get things done for you as well. With TUIs I think either these are hobby projects or meant for people who are obsessed with speed.

Though speed impacts are also something which I am uncertain about. Comparing Vim with IDEs, for sure there will be few things which are faster in vim but decent no of things which can be done faster in an IDE as well, so can't comment on your overall speed gains.

Amazon used to be really customer centric 5-10 years ago, I remember once I ordered a physical book which was late in delivery and I urgently needed that book, so they gave me a free kindle edition till the book got delivered.

or get requests with query params already handles this in majority of the cases, unless the query size is too big (which ideally should not be the case since in the end it is a get request)

This theory seems to be BS, If let us say a founder is raising a seed round of 2m at 20m valuation, then according to hypothetical accrual tax rate, they would need to pay a tax of ~ 3-4m.

Google is successful and it's page rank algorithm also does not deliver correct results all the times.

Yes if Clippy was released in 2025, it would surely be stealing your data without thinking twice.

Is it ? In gambling your odds are fixed, but in real life, wouldn't you get better at solving problems with each iteration ?

I think there is one counter argument, LLMs are speeding up everything, including the speed of learning, which also implies that companies that might have bad processes would learn and move to good processes as well on the way.

Example, one of many things, in our SDLC process, now we have test cases and documentation which never existed before (coming from a startup).

Probably the OP might have used LLM's to structure the document, but the sentiment `Automating stupidity = faster stupidity` is a great take away.

If having a strong SEO strategy was worth 100 points two years ago, how many points would you assign to it in today’s environment? Has its importance increased, decreased, or shifted in nature ?

Understood make sense

Reminds me of the Apple vs Pear law suit

https://www.entrepreneur.com/business-news/apple-sues-small-...

The dispute was settled because Pear agreed to slightly alter its logo, instead of continuing full litigation (maybe because of resources / dollars it would consume)

It has started to backfire.

Claude also had to a pay almost 1.5b for illegally training / scrapping.

https://www.cnn.com/2025/09/05/business/anthropic-ai-settlem...

There will be chaos, all code bases usually have tons of side effects.