Andreas Eisele

🏢 Developer @JetBrains
🍃 vegan
🐈 crazy cat person

[Andreas Eisele @ae____]

TIL: range checks in Kotlin. For the odd case every now and then when you need to know whether a value is in a range. https://t.co/RquUo6B1iH

[Andreas Eisele @ae____]

Imagine buying a first class plane ticket but the airline doesn't like your pants, so they decide you're traveling by train instead while keeping the ticket price. That business literally wouldn't fly. But somehow the rules around Fable are fair game for everyone.

[Andreas Eisele @ae____]

For the last weeks I've been using the "superpowers" set of skills with Claude and Codex. Honestly though, I'm not sure if it costs me more tokens than it brings value. How would someone gauge that short of running a costly track of benchmark evaluations?

[Andreas Eisele @ae____]

It's superlative season again folks. If you haven't blown your AI budget in a single Claude Fable invocation by now, you're not gonna make it. Also every other model is now unusably stupid, not worth your effort. Know the rules!

[IntelliJ IDEA, a JetBrains IDE @intellijidea]

“Functions and objects should never mix.” – That was the dogma before Scala. @odersky joins Marco to talk about creating Scala, influencing Java/Kotlin/Swift, AI-generated code, capabilities, and why human code review may not scale anymore.
🎥 Watch now: https://t.co/ACIXHlwjEL

[Andreas Eisele @ae____]

It's interesting to see how oblivious GitHub Copilot users were to token usage. They lived in this convenient world of per-session billing. Suddenly the thing that their org is paying for anyway doesn't look like such an obvious deal anymore.

[JetBrains @jetbrains]

Mellum started with code completion.
Mellum2 is built for more – handling both natural language and code.
A 12B-parameter open-source LLM for routing, RAG, and sub-agents, optimized for ultra-low-latency inference.
Now on @huggingface.
Learn more: https://t.co/28sG8Ql52L

[Andreas Eisele @ae____]

Even Karpathy said in some recent interview that generated code usually has strange brittle abstractions, is full of copy paste and making stuff simpler is still an unsolved problem. At the same time LLMs are great at doing reviews. Maybe we should concentrate on that more?

[Siva @sivalabs]

I just want to log some value temporarily for debugging purpose. Can I do it without adding log statements in code?

YES.

While I am already running the app?

YES.

Without restarting the app?

YES.

Even in 3rd party code?

YES.

https://t.co/EMfGRXOzM1

[Andreas Eisele @ae____]

I have neither the budget nor the quota to verify the new "dynamic workflows" feature in claude code. But I think it demonstrates the highly uneven playing field in software development. The few companies that operate on zero token costs can iterate immensely fast now.

[JetBrains @jetbrains]

We did it again.

Something new for the open-source LLM space is coming to @huggingface soon.

Stay tuned.

[Andreas Eisele @ae____]

Even my spam mails now promise me cheap AI tokens instead of millions of dollars from a deceased prince.

[Andreas Eisele @ae____]

Running a big task with sub agents in claude is wild. I now have "implementor" agents gaslighting the "reviewer" agents that their implementation is fine despite concerns.

[Andreas Eisele @ae____]

Token cost economics mean either local / at premise models and hardware become a thing or human programmers keep their jobs after all. Continuously 3x-ing the bills is not viable.

[Andreas Eisele @ae____]

@sivalabs Don't take away my "well, actually..." talking points from me.

[Andreas Eisele @ae____]

Help me please. I'm somehow missing the picture. Token costs are skyrocketing. At the same time AI is gonna replace all humans because it's so much cheaper? I don't get it.

[IntelliJ IDEA, a JetBrains IDE @intellijidea]

Don’t add breakpoints and suspend your @springboot app just to see what's actually injected. Keep your flow intact! With #SpringDebugger in @intellijidea, live bean implementations appear right in your code via inlays. Get full runtime visibility with zero pauses! https://t.co/6fQuyQVEWk

[Siva @sivalabs]

Oh wow.

Now people are more concerned about the feelings of LLMs.

You can fire 1000s of employees without thinking for a second, but you better not hurt the LLMs feelings.

I think we crossed all levels of insanity 😔

[Andreas Eisele @ae____]

Sure, nobody writes code anymore. But do you still delete code? Imagine paying for multiple AI agent round trips instead of just using the "safe delete" action in IntelliJ.