poojan

no

[Jesse @jesse_vermeulen]

finally cracked the code for liquid glass on the web

[Abhishek B R @abhitwt]

pov: you visit linkedin for the first time ever:

[Shravani @shrav_10]

I wasn't even aware that microsoft(dot)ai existed.

There's barely any mention of it, and it doesn't even have a distinct logo. Was it intentionally kept under the radar, or have I just been living under a rock?

[poojan @pizzacubess]

@Olacabs approx a year ago, charged 900 instead of 450 for one of my rides, and just said voucher issued with no option to reach out to the ola customer care, i still don't know where my money is, and how do I recover it, expected from @Olacabs , stop using this pathetic service

[Kristel @Web3Kristel]

@OpenAI Aside from this being a terrible move, the timing of this announcement tells me the team is not reading the room.

Your competitors are taking your users right now. This is not the time to piss off your base even more.

[poojan @pizzacubess]

@MargauxxBillard @t_winkworth Iconic picture!

[Margaux Billard @MargauxxBillard]

@t_winkworth

[BLAZE @BLAZETVUK]

It's true that @Rbargainhunters talks the talk, but can he walk the walk against @MaryPadian? #StorageWars

[Santiago @svpino]

How to detect hallucinations when evaluating your LLM.

This is a fantastic article (link in the next tweet) with step-by-step instructions on how to use an LLM-as-a-judge approach to detect hallucinations.

Here is a quick summary:

1. You'll use an LLM to judge outputs
2. You'll use a prompt (see image) that will generate a metric
3. You'll use structured outputs (for example, JSON)
4. You'll build an inference pipeline
5. You'll set up your evaluation process

At a high level, the idea is to produce your outputs and run them through the judge to determine whether there are hallucinations or not.

The key here is to do this quickly (keeping latency to a minimum), reliably (we need to trust the evaluation results), and as simply as possible (we need a way to orchestrate all of this.)

The article uses Opik (an open-source evaluation library) to orchestrate this process. There's a Colab with all the code you need.

[Fred Bear Bulldog™️ @BulldogTCOB]

your Oxford, MS tailgater/booster couple

[AF Post @AFpost]

X user Lilly (@llddiiss) is going viral on X and TikTok for calling Jewish influence on politics “gay," unapologetically using the N-word and saying she’s America First.

She’s since thanked “the black community” for launching her “new career in conservative media.”

Follow: @AFpost

[Political Point @World_News8888]

‼️White girl uses the N word in a video and she becomes famous overnight 😳 @llddiiss

[Lilly @thelillygaddis]

#1A #response

[locdinwithzaza @locdinwithzaza_]

S/o to my nigga @llddiiss ❤️

[Prost Stan Account 🐝 @f1_headassery]

Good morning classic F1 Twitter

[Priyanka Lahiri @PriyankaLahiri_]

Packaged juices will always be high on carbohydrate whether you add sugar or not.

1. Its always important to check packaged food labels

2. Avoid fruit juices and go for whole fruits instead.

3. Juice concentrates usually have sugar added while processing + juice itself is fructose.

[Anugya Jha @AnugyaJha3]

How much should a customer endure the scams of the food industry by the misuse of language?

This 100% pomegranate juice with no added sugar sounded relishing while we ordered it. But upon turning to its ingredients we were horrored by its cunning wordplay.

[Larry Leprechaun @LarryLeprechau]

Congrats to Team USA who only lost by one point 111-110 to Team India in the World Series Cricket competition happening in New York.

[Charli @charli_xcx]

i am extremely proud of my indian heritage. i love my roots & my family. don't call me/anyone "dirty" bcoz of the colour of someone's skin.

[Remi @4unclelala]

drivers reactions to lando first win- a thread