Saurabh Kumar ✓

@drummatick · Tokyo · joined 04 Sep 2016

e/local-llm 🚀 Building a local Qwen Harness LLM fine tuning | Harness optimization | Evals and Memory systems for Agents (opinions my own)

21 308Followers

561Following

21 763Posts total

833.5KViews on collected posts

Últimas publicaciones

Saurabh Kumar ✓@drummatick21 Sep

JEV is 8-10% more likely to commit an error than GPT-6 on decision making. Sounds low, right? its not. Each wrong model pick, each wrong reasoning level pick will pile up that error on top of each other in a long horizon task Will you save some money? yes 205 views · 0 likes · 0 reposts · 0 replies Open on X →

Saurabh Kumar ✓@drummatick21 Sep

bruhhhh 336 views · 1 likes · 0 reposts · 0 replies Open on X →

Saurabh Kumar ✓@drummatick21 Sep

give or take 2 days, some tech bro is gonna reinvent RAG using JEV and gonna call it revolutionary (they don't know what embedding search is) 253 views · 3 likes · 0 reposts · 0 replies Open on X →

Saurabh Kumar ✓@drummatick21 Sep

if you're still confused if JEV is for you, where to use and how to use it, do read this article! 229 views · 4 likes · 1 reposts · 0 replies Open on X →

Saurabh Kumar ✓@drummatick21 Sep

People use Jev to pick a model before a task. I made it change GPT-6's reasoning effort inside Codex DURING the task. More thinking when stuck. Less for routine steps. 50% lower Astra costs in my tests. Faster runs, without breaking prompt caching. https://t.co/NrpGFfWW9v

0:23

55.2K views · 811 likes · 17 reposts · 58 replies Open on X →

Saurabh Kumar ✓@drummatick20 Sep

Does JEV really save the cost or does it come with a trade-off? I ran the experiment on the Banking77 dataset on Hugging Face. The results? GPT -5 beat JEV by 3.2%, but at what cost? 32X. Yes, 32X I also ran a cascade analysis(JEV + GPT5), check it out!! https://t.co/EIZPoJ7OlF 958 views · 8 likes · 4 reposts · 6 replies Open on X →

Saurabh Kumar ✓@drummatick26 Nov 2023

@drummatick Can you confirm the playlist for Linear Algebra https://t.co/WAXm5iNGnG 12.3K views · 36 likes · 2 reposts · 1 replies Open on X →

Saurabh Kumar ✓@drummatick25 Nov 2023

330 likes and 635 bookmarks Bookmarks twice the amount of likes Meaning people think it’s useful (Please consider sharing it as well :p) 16.2K views · 48 likes · 0 reposts · 0 replies Open on X →

Saurabh Kumar ✓@drummatick25 Nov 2023

One thing I forgot to mention is that, apart from these openly available resources I took university courses on ML, deep learning and information theory as well 21.6K views · 75 likes · 2 reposts · 2 replies Open on X →

Saurabh Kumar ✓@drummatick25 Nov 2023

I know this might get more bookmarks than likes, but do share it :’) 30.2K views · 88 likes · 1 reposts · 6 replies Open on X →

Saurabh Kumar ✓@drummatick25 Nov 2023

This tweet is about how I have studied ML and made it my profession. I'll share the resources I've used and the sequence of my study. Straight to point ML pre-requisites(maths) : Linear Algebra, Probability Theory, Calculus, Optimization Theory(optional), Information 696K views · 4.1K likes · 559 reposts · 81 replies Open on X →

Frente a cuentas del mismo tamaño

6 publicaciones de los últimos 90 días, junto al rango de 10K–100K seguidores. llega a menos gente que cuentas de su tamaño.

Visualizaciones medianas294esta cuenta924mediana de 10K–100K

Alcance, %1.38%esta cuenta3.62%mediana de 10K–100K

Interacción, %1.40%esta cuenta1.52%mediana de 10K–100K

Métrica	Esta cuenta	Mediana de 10K–100K	Proporción
Visualizaciones medianas por publicación	294	924	0.32×
Alcance (visualizaciones ÷ seguidores)	1.38%	3.62%	0.38×
Tasa de interacción	1.40%	1.52%	0.92×

Otras cuentas de este rango → Comparar con otra cuenta → Cómo se construyen estas referencias →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

696K25 Nov

30.2K

21.6K

16.2K

12.3K26 Nov

95820 Sep

55.2K21 Sep

229

253

336

205

Last 11 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.68%25 Nov

0.31%

0.37%

0.30%

0.32%26 Nov

1.98%20 Sep

1.62%21 Sep

2.18%

1.19%

0.30%

0.00%

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.

What the audience does

Likes32.6%5 146 in total

Reposts3.7%586 in total

Replies1.0%154 in total

Quotes0.3%45 in total

Bookmarks62.4%9 838 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Últimas publicaciones

Frente a cuentas del mismo tamaño

Growth & engagement

Views per post

Engagement rate per post

What the audience does

Cuentas similares