RavenUmbra

@This_is_Ramana @om_patel5 I think you never compared the coding quality of them did you?

@RavenUmbra @om_patel5 As a developer he is supposed to know what model to use for what purpose It’s like bringing a JCB to lay a brick 🤷🏻‍♂️ Haiku for scraping Sonnet for coding and other basic stuff Opus for deep research with documents

saying "hello" to Claude on the Pro plan now costs 2% of your entire session usage one message. "hello, how are you?" that's it. this is why people are mass migrating to Codex right now because its literally impossible to reach limits anthropic needs to fix this before they lose the crazy amount of developers they just gained

@NalaRubi @Stahl_ratte @temberaubend Und bei Schreibmaschinen seid ihr auch stehen geblieben. Aber nicht aufregen, eure Rente zahlen wir ja brav. Generation nach mir die Sintflut. Niemand hätte es im Leben so einfach und billig wie ihr.

@RavenUmbra @Stahl_ratte @temberaubend PDF? Bist meschugge? Wir haben damals auf Schreibmaschinen, möglichst mit Pauspapier drei Durchschläge gemacht. Bei Tippfehler von vorne angefangen. So einfach war das damals auch nicht. Heute kommt man wohl einfacher zu Geld als damals.

Wie schafft man es eigentlich mit einem 40h Vollzeitjob (mit Pendelzeit obendrauf) Haushalt, Essen kochen, Sport machen, Freunde, Familie, ggf. Partner treffen, Hobbys nachgehen und mal zur Ruhe zu kommen, unter einen Hut zu kriegen? Frage für einen Freund

@lydiahallie I thought you just run AI agents and they build and fix everything? Wasn’t that the latest PR stunt? That’s now a very long running issue with no fix in sight

Update: still working on this. It's the top priority for the team, I know this is blocking a lot of you. More as soon as we have it.

We're aware people are hitting usage limits in Claude Code way faster than expected. Actively investigating, will share more when we have an update!

@nopicklespleaze @lydiahallie We are fan boying buggy rate limiting now? Oh dear

@ishuagra02 @om_patel5 @Anthropic Caching has a 5 min TTL. I don’t think this is the issue, assuming your ever take more than 5 min between interactions.

@om_patel5 There seems to a cache invalidation bug that's leading to higher token consumption. Here's workarounds which reduced my usage: - Use Claude Code via: npx @anthropic-ai/claude-code - Do NOT use --resume for chats, this will break caching.

@This_is_Ramana @om_patel5 Literally their first sentence in the announcement: „The new Claude Opus 4.6 improves on its predecessor’s coding skills.“. They don’t market it at all for „deep research (only)“

@om_patel5 You are not supposed to use Opus except for deep research purpose 🤦🏻‍♂️

@summer_mini @Valve2711 @temberaubend Ja dann würde ich an deine Stelle aufhören zu projizieren. Wir geben ja Gas für deine Rente von der wir nicht sehen werden. Mehr raus als rein ist ja das Motto von euch seit Geburt 😊.

@RavenUmbra @Valve2711 @temberaubend Ja, es ist schon einfach, wenn man seine eigene Unzulänglichkeit auf andere projeziert. Mein Tipp: Aufhören zu heulen und einfach mal Gas geben. Von der Rente bin ich übrigens noch Meilenweit entfernt und mein Eingenheim mit >200qm bekommt man leider auch nicht für 3,50€ 😉

@summer_mini @Valve2711 @temberaubend Wieder ein Boomer mit 3.50€ Miete und fetter Rente du nicht mal ne Maus mit dem PC verbinden kann und von Belastbarkeit schwafelt. Deine Produktivität liegt sicher bei fast 0

@Valve2711 @temberaubend Das Problem ist, dass die jungen Leute heute oft nicht mehr belastbar und noch dazu absolut unorganisiert sind. Da kann einen so ein 40 Std. Job dann schon ganz schön aus der Bahn werfen. Jeder möchte doch nur noch möglichst wenig arbeiten aber maximal verdienen. Finde den Fehler

@Stahl_ratte @temberaubend Bei euch war die Miete 3 Euro 50. Und eure Produktivität. Naja. Als PDF? Lass mich das mal ausdrucken! Einfacher Geld machen als zu dieser Zeit ging ja nicht. Und so wenig gemacht dass alles zerfällt

@temberaubend Warum hat man es früher geschafft bei einem viel zeitintensiveren Alltag in einer Offlinewelt: Kein Internet-Shopping, ohne soziale Medien, keinerlei Onlinetools, usw?! Bin in den 70ern geboren und lache mich krumm über das heutige Gejammer! 😂

@temberaubend Shareholder-Value ist nicht dein Hobby? Ich sehe da Optimierungspotential!

@garrytan Publish the code and real engineers can easily point out what’s wrong with it.

Absolutely insane week for agentic engineering 37K LOC per day across 5 projects Still speeding up

@shiri_shh Claude code isnt complex software though. Heck no one even cares if it has bugs

Creator and head of Claude Code: "100% of my code is written by Claude Code. I have not edited a single line by hand since November. Every day I ship 10, 20, 30 PRs… I have five agents running while we’re recording this."

@NikoStarus @Layton_Gott And thanks to guys like you critical systems crash because you don’t grasp memory safety, concurrency or any other thing beyond „lgtm“. Wether you like it or not, you can’t judge it. You think it looks good, but you don’t know it. In fact you are a human LLM

@RavenUmbra @Layton_Gott coding isn’t the same skill as judging what the ai does. systems engineering and orchestration is the new meta whether you like it or not

I CAN'T code without AI. And I'm not even a little ashamed to admit that. "But could you build it without AI?" Who cares. We HAVE AI. That question is like asking a pilot if they could fly without a plane. I've shipped more with Claude Code in 3 months than most "real developers" ship in a year. Users DON'T care how you wrote the code. They care if it works. No matter if that's with AI or manual.

@NikoStarus @Layton_Gott He said he can’t code without AI. Hence he can’t judge what the AI does. Like putting anyone in the cockpit because there is an autopilot. Would you enter that plane?

@Layton_Gott the pilot analogy is perfect. nobody asks a surgeon if they could operate without instruments. AI is just the instrument now. the skill is knowing what to build, how to architect it, and when to ship. that never changed

@P33RL3SS @successfulmen_t @om_patel5 You seem to write simple code and architecture then if you haven’t encountered LLM limits.

@successfulmen_t @om_patel5 What does that have to do with if the code was generated or written? I can say the name of the algorithm, the style (functional), the principles (SOLID), the complexity low, and the LLM generates according to my specification. So why does it matter anymore? The LLMs are good now.

THE ANTHROPIC TEAM DOESN'T WRITE CODE ANYMORE. this guy's friend got hired at Anthropic 3 weeks ago. nobody on his team has hand written code in months. they run multiple agents in parallel and act more like managers than engineers. his friend said if you're just watching an agent code, you're already behind that idle time should be spent spinning up another agent and directing it somewhere else. the point is that the new method isn't "use AI to code faster." it's "you are the product manager, the agents are your engineers, and your job is to keep all of them running at all times"

@GrmlBob @successfulmen_t @om_patel5 I doubt they keep context. The people who work like that at my place don’t review enough/ understand it and ship garbage

@successfulmen_t @om_patel5 And the take is bullshit, because no human can keep up with this much context, context switch and so on. These people would be burned in weeks.

@kckrish007 @successfulmen_t @om_patel5 Have you ever coded efficient code? If you see what AI produces it’s not efficient at all most of the time. I’m doubting your ability to judge itto tbh

@successfulmen_t @om_patel5 There are no best practices when agents code. When we used to code by hand, we sometimes dumb down the code to make it human readable. It's a tradeoff. But the agents write most efficient code which also means, it takes more than few seconds to understand what it wrote.

@Dgkalx1267 @successfulmen_t @om_patel5 You are one of the guys publishing api keys in frontend for sure

@successfulmen_t @om_patel5 Your Jr devs are not using AI right then. I've never written a single line of code my entire life. I'm already a sr. dev pushing out production ready apps in weeks. Set the environment, set the guardrails, provide context, and let the AI cook. This is the the way.

@0xdippo @successfulmen_t @om_patel5 If you can merge as is your software is simple, architecture doesn’t matter and it will simply be unmaintainable in the future. If you think AI produces quality, you are a bad engineer. Sorry

@successfulmen_t @om_patel5 It’s been confirmed by the team repeatedly. Sorry your team is lagging.

@om_patel5 Anthropics software is one where architecture, consistency, high availability etc doesn’t matter though. Above ML they don’t have any complexity