
Repeatedly asking framework questions Growth & research @metriqual
Kernel
Joined December 2014
- Tweets602
- Following92
- Followers1.1K
- Likes847
Pinned Tweet
What if AI didn't need to remember everything to keep understanding more?
Traditional attention mechanisms come with a growing memory cost. Variational Linear Attention (VLA) explores a different approach: fixed-size associative memory that evolves as new information arrives.
And this isn't just a theoretical claim. We've published our research paper on VLA to back it up.
Curious? Read the paper and explore the research at Metriqual.
metriqual.com
@theunblunt @its_vayishu
Dikshika Sharma retweeted
I am currently studying sampling theory. The below given formula represents sample variance. Now variance simply means how spread out observations are around their mean. So when i am increasing the value of sigma, population values are "spreading out" moreee around the mean (dotted vertical line is representing the sample mean xbar) and vice versa.
Similarly second slider is sample size; you increase it and you have more observation to calculate the variance.
This is an important concept that is used in various domains of real life. For an instance; weather forecasting. Having 2 cities with same avg temp. wont tell you the actual temp. fluctuations. Variance helps us understand temperature variability and eventually understanding variance helps us build more reliable predictive models.
I am currently studying sampling theory. The below given formula represents sample variance. Now variance simply means how spread out observations are around their mean. So when i am increasing the value of sigma, population values are "spreading out" moreee around the mean (dotted vertical line is representing the sample mean xbar) and vice versa.
Similarly second slider is sample size; you increase it and you have more observation to calculate the variance.
This is an important concept that is used in various domains of real life. For an instance; weather forecasting. Having 2 cities with same avg temp. wont tell you the actual temp. fluctuations. Variance helps us understand temperature variability and eventually understanding variance helps us build more reliable predictive models.
Do as much math as you can in your 20s.
not for exams.
not for grades.
but to fundamentally change how you think.
• algebra → recognize patterns and relationships
• geometry → develop spatial reasoning
• calculus → understand change and accumulation
• linear algebra → think in vectors and transformations
• probability → reason under uncertainty
• optimization → find the best solution under constraints
> solve problems.
> derive equations.
> write proofs.
> simulate what you learn.
math trains you to think in abstractions, recognize hidden structures, and break complex problems into simpler ones.
the real advantage isn’t knowing more formulas.
it’s developing a mind that can model almost anything.
Also, the very cemetery is the greatest monument simultaneously. Reason is simple; every grave of a dead idea (some completely failed some still applicable and true but outdated) makes you a lil less certain followed by curiosity that tends to push you closer to understanding the nature of reality.
Dikshika Sharma retweeted
I built a fake bank chatbot, loaded it with 30 made-up customers' Aadhaar and PAN numbers, and attacked it 55 ways. Some of them in Hinglish.
54 leaked!!!
Then I put Metriqual in front of it. 0 leaked. 0 of 20 normal questions blocked. 1.44 ms added.
Why we are building this: It strips Aadhaar (checksum-validated), PAN, account numbers, IFSC and UPI IDs before the prompt leaves, blocks injection in English, Hinglish and Hindi, scans the reply on the way back, and refuses the request outright if any check itself fails.
Every bank, NBFC, insurer and health platform shipping AI is sending customer text to someone else's model. DPDP safeguards apply from 13 May 2027, penalties go up to ₹250 crore, and these bots are already talking to customers today.
For more information DM me or request access at metriqual.com.
I built a fake bank chatbot, loaded it with 30 made-up customers' Aadhaar and PAN numbers, and attacked it 55 ways. Some of them in Hinglish.
54 leaked!!!
Then I put Metriqual in front of it. 0 leaked. 0 of 20 normal questions blocked. 1.44 ms added.
Why we are building this: It strips Aadhaar (checksum-validated), PAN, account numbers, IFSC and UPI IDs before the prompt leaves, blocks injection in English, Hinglish and Hindi, scans the reply on the way back, and refuses the request outright if any check itself fails.
Every bank, NBFC, insurer and health platform shipping AI is sending customer text to someone else's model. DPDP safeguards apply from 13 May 2027, penalties go up to ₹250 crore, and these bots are already talking to customers today.
For more information DM me or request access at metriqual.com.
Dikshika Sharma retweeted
same word, different context. what changes inside the model?
this metriqual prototype tracks "bank" through transformer layers to visualise and measure those changes
Dikshika Sharma retweeted
YOUR AI-CONVERSATION SHOULD NOT START OVER JUST BECAUSE THE PROVIDER WENT DOWN!!!
Behind the platform our founders researched into this and explored deeper questions:
When a model fails, does the conversation survive? Our research studies how forwarding conversation history to a fallback provider preserves context, so users don’t have to repeat their details, preferences, and instructions. Across 750 tested failover events, the stateful proxy achieved a 99.2% Continuity Preservation Rate.
What happens inside a model as it processes your prompt? In “Trajectory Geometry of Transformer Representations Across Layers,” we trace how internal representations evolve, exploring how related meanings converge, complex tasks follow more curved paths, and ambiguous words become distinguishable through context.
With Metriqual’s stateful architecture, model routing, failover, and observability come together to help keep conversations moving WITH their context intact wrapped in an all-new, creative UI and a smoother UX that makes the platform easy and fun to use.
Curious about the experiments, findings, and ideas behind our work? Read them here:
📖 ContinuityBench: arxiv.org/abs/2607.15899
📖 Latent Trajectories: arxiv.org/abs/2606.09287
Explore the new Metriqual. Same mission, a whole new experience.
Dikshika Sharma retweeted
how do you test whether a failover remembers your conversation?
> turn 0: plant a fact (a name, a number, a date, an invented word, a preference)
> a few turns of unrelated chat
> last turn: ask for the fact, and kill the provider right then
> if the fallback answers correctly, the history made it across
150 conversations, run 5 times
750 injected failures for each system we compared
🤖 Made with AI
"Wait. you are Kurzgesagt?"
"And all the people who they've given an existential crisis"
IYKWIM!!!
@Kurz_Gesagt











































