rikko

SWE MSc 🤙🏻

[rikko @rikkoo0]

@antoniosarosi i took a bit to write the second message....
https://t.co/4jwFgmz9NJ

[rikko @rikkoo0]

@antoniosarosi LLMs spit on that and write whatever fits just to reach a goal, taking that idea of passion, intent and longevity and discarding it.Also the community, it created deep discussions about ideas and user needs. Ultimately how do you transition from technical to a focus on business?

[rikko @rikkoo0]

@antoniosarosi LLMs spit on that and write whatever fits just to reach a goal, taking that idea of passion, intent and longevity and discarding it.Also the community, it created deep discussions about ideas and user needs. Ultimately how do you transition from technical to a focus on business?

[rikko @rikkoo0]

@antoniosarosi I dont think its that shallow. Personally software was always about solving problems because they looked challenging and limitless. They require different patterns and algorithms that would fit the problem, the user experience is one of them.

[rikko @rikkoo0]

@addyosmani Theres a high chance you need to make AI look good because you still own google stock. You're just a shovel seller.....

[rikko @rikkoo0]

@Love2Code Anything that doesnt change your output is a valid use of LLMs IMO.

[rikko @rikkoo0]

@maxbittker I wonder how recent LLMs would perform in the game Toribash, its a game that requires really good Spatial reasoning

[rikko @rikkoo0]

@TheRightColumn @KylerJohnsonDev When you build something you need to be able to maintain it, AI generation doesnt count as building. You're missunderstanding his point as a "gotcha" moment to get a reaction. In 3 years i havent seen a person that went from layman to a hired SWE only using LLMs.

[rikko @rikkoo0]

@TheRightColumn @KylerJohnsonDev It can walk a non technical person through a below-average project that will fold under load. Englighten me, whats the difference between that and tutorial from the pre-ai era? Wheres the innovation here? LLMs are as a substitution for guidance as a tutorial for a non-technical

[rikko @rikkoo0]

@burkeholland it will never get good enough where you wont have to read it at all, even compilers produce security vulnerabilities sometimes. 1/10 weak bait

[rikko @rikkoo0]

@jacob_bacus @sickdotdev The alternative would be to dust off the good old SPARQL. The elephant in the room is the websocket, they tried to improve it but it will take a while until every service is using the new MCP 2026-07-28.

[rikko @rikkoo0]

@janetacarr HAHAHAHAHAAHAHAHA

[rikko @rikkoo0]

@VictorTaelin You have no idea how i was hoping did this, thank you.

[rikko @rikkoo0]

@arvidkahl @XCSme "you're holding it wrong" ahh response

[rikko @rikkoo0]

@hagia_machina The mark can be easily broken by changing the content, you know? like y'all claim to "verify", edit and rewrite...:) It might be a good time to start using AI as feedback instead of a cheating tool tho

[rikko @rikkoo0]

@Sams_Antics @dieaud91 you should shame people that use AI for most of their work. Its a good article, but being nitpicky about a few cases where AI detectors failed without seeing the accuracy of the new ones is a bit disingenuous, In the same level of trashtalking LLMs while being on the free plan

[rikko @rikkoo0]

@SynorosHQ @dieaud91 right, are you going to pass through that model every single piece of text you generate? that sounds realistic for a day max

[rikko @rikkoo0]

@kelvinbuildss i think its more because its a deterministic way to evaluate how people navigate through problems at scale. Architectural questions are subjective and harder to qualify, I'm sure software architect interviews are mostly the way you're describing

[rikko @rikkoo0]

@frizankthetank @TheMG3D then rewrite some parts and improve it like y'all claim to do all the time, no?

[rikko @rikkoo0]

@MarikoSanogo @TheMG3D you can always take feedback and edit it yourself. If you need to copy paste the whole thing from claude then you're the co-creator.

[rikko @rikkoo0]

@haider1 @andrewho03 creativity is the ability to take unrelated ideas and implementing them, no LLM will take the rust error message structure and implement it as a custom error handler in another language because its more pleasant to work with, for example. Their scope is incredibly narrow