Aldrin

It is "piques your interest" NOT "peaks your interest" !

@milesmintIN @virsanghvi Similar case with icici bank. They either reply with a copy paste template or do not bother to reply at all

HDFC Bank seriously needs to rethink its Relationship Manager model. It feels like fresh graduates are being assigned as RMs without proper training on HDFC products and internal processes. In my 20+ years as an HDFC Imperia customer, I’ve rarely received meaningful support from an RM: • Either they don’t have authority • Or they don’t want to take ownership Simple requests that should take 5 minutes end up dragging for a week. Even escalations rarely help. In comparison, ICICI Bank RMs are far more proactive — they at least try to help, even when something is technically outside their scope. Sadly, HDFC seems to be growing more arrogant with time, forgetting that premium relationships are built on service, not branding. 📌 This post is based on my personal experience. #HDFCBank #CustomerExperience #Banking #Imperia

If you are not generating code in assembly then are you even vibe coding ? 😀

@NanoBanana different coloured and type of cats forming a tower of babel soaring into the sky

Check out my latest article: Cypher: Don't just generate, Iterate! linkedin.com/pulse/cypher-d… via @LinkedIn

Calling Python from a Java AI framework linkedin.com/pulse/calling-…

The plugin reviews are not great plugins.jetbrains.com/plugin/24198-g…

Gemini code assist in IntelliJ is so bad that I prefer to use Cursor for code gen and only come into IntelliJ to debug or code some specific business logic. I'm using IntelliJ less and less

@karpathy Updating my LinkedIn to "Context Whisperer" 😉

+1 for "context engineering" over "prompt engineering". People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window with just the right information for the next step. Science because doing this right involves task descriptions and explanations, few shot examples, RAG, related (possibly multimodal) data, tools, state and history, compacting... Too little or of the wrong form and the LLM doesn't have the right context for optimal performance. Too much or too irrelevant and the LLM costs might go up and performance might come down. Doing this well is highly non-trivial. And art because of the guiding intuition around LLM psychology of people spirits. On top of context engineering itself, an LLM app has to: - break up problems just right into control flows - pack the context windows just right - dispatch calls to LLMs of the right kind and capability - handle generation-verification UIUX flows - a lot more - guardrails, security, evals, parallelism, prefetching, ... So context engineering is just one small piece of an emerging thick layer of non-trivial software that coordinates individual LLM calls (and a lot more) into full LLM apps. The term "ChatGPT wrapper" is tired and really, really wrong.

Java Duke Nukem! jetbrains.com/lp/java-30/duk…

What are hackathons like in the "vibe coding age" ?

@sundarpichai Once you add auth, streaming etc to MCP, it will soon look similar to existing ones

@sundarpichai MCP is smart in the sense that it puts the onus of support on the developer of that service. However imho, it would be much better if LLMs can grok the various protocols already available and handle the communication themselves.

To MCP or not to MCP, that's the question. Lmk in comments

@iam_preethi It's amazing how these LLMs all think alike and speak alike

I have been interviewing for various roles at Ferta. One step of our process is a take-home assignment, and almost everyone uses ChatGPT for written tasks these days. It sounds "good", but is so generic and indifferent. Reading something that was written by a human vs. AI is immediately obvious. When it written by a human, you can hear the person at the other end talking to you via their writing. Not the case with AI.

@tagir_valeev It makes up hyperlinks. I've never got a valid link ever. The domain is usually good, but the rest is just hallucinated

I was searching for an XKCD comic, so I decided to try ChatGPT instead of Google. It provided a link, which I checked, and it's definitely a wrong comic. I tried asking again but it was insisting. After longer discussion and confusion, it appeared that the comic is correct (it's 'Success' and its number is 349) but the URL is wrong (it leads to 'Fixing Problems', #1739). I still don’t get used to the tool, which, while being helpful, may make mistakes in things that I consider trivial for computers. E.g., if you are speaking about A, why are you linking the page about B, provided that both pages are static and exist for many years? Something that I assume as granted (when search engines provide a link to a page, they describe the same page) doesn't work robustly anymore.

meanwhile, European power sockets

Bryan Adams totally rocked last night 🤟

Até logo, Lisboa

@InAbuDhabi @coldplay Waiting for the third show 😃

ABU DHABI 📣 @coldplay just announced a second show #InAbuDhabi on Jan 12 at Zayed Sports City. Ticketing link: ticketmaster.ae/artist/coldpla…