← Back to leaderboard
archiexzzz

Archie Sengupta

Verifiedactive
@archiexzzz

researcher @midcenturyai (robotics & simulation) // researcher @sarvamai // pre & post-training // prev: @atomicworkhq, 5x yc dev

RosterHumor & Meme AccountsTech Career & Coding Education
69,762 followers · 1,200 following · Last polled 1h ago
Perf. Score
74.1
Rank #45
Value Score
36.1
#179 of 273 priced
Median views
8.8K
p25 4.2K · trailing 7d
Followers
70K
+0.1% 7d
ER (impr.)
3.91%
eng ÷ impressions
ER (foll.)
0.65%
eng ÷ followers
Est. CPM
$79.62
QT-rate basis
WoW views
-47.7%
▼ falling

Rates & value · est. from 8.8K median organic views

overpriced
Quote tweetbasis
$700
$79.62 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 36.1 · rank #179 of 273 priced$2788.84 / 1K engagementsrate age unknown (pre-audit import)

Trends

Follower growth
Median views / day
Engagement rate
View distribution · trailing-7d posts

Each blue dot is one organic post. The Performance Score ranks by the median (8.8K), not the mean (12K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (4.2K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
I seriously don’t understand why Elon hasn’t realized that the average Indian pays only about $40/year for unlimited 5G internet vs Starlink’s $1,000/yr + hardware kit costs.frozen13d ago137K2.3K8614418761.92%
elon musk needs a better PR agency bro is getting bashed left, right and centre 😅frozen14d ago40K4.0K2277253710.68%
As someone who has hacked countless companies and received thousands of dollars in bounties, let me tell you this: the future of cybersecurity is genuinely scary if companies, governments, and international bodies do not take it seriously. This week’s incidents w/ eMudhra, TiDB/PingCAP, PHP RCE, and the SBI data leak were just a trailer for the very dangerous future we are heading toward. Soon, we may see cases where a city’s electric grid, water-supply systems, and physical infrastructure are sabotaged through cyberattacks carried out by adversaries.
23h ago
35K
1.4K
61
37
11
253
4.98%
for the young folks out there: be humble. humility is the greatest virtue. skills will come and go; your failures and successes will come and go. what always remains constant is how you treat people and how you make them feel. i hope the world becomes a little kinder.frozen29d ago20K1.3K9824101627.89%
when in doubt, just reverse engineer https://t.co/UtbOykVkKsfrozen18d ago19K4653511343.14%
it took 7 years!!!!frozen7d ago18K266631221.70%
everyone should have their own software factory engine. use: > brew tap manaflow-ai/cmux > brew install --cask cmux https://t.co/igOzzkEOBbfrozen18d ago17K2282631932.60%
ok guys i have attained clarity in life today. real screenshot btw, not ai :)) https://t.co/wc3OMH4pF4frozen11d ago16K3971210282.78%
did anyone distill gpt-6.1-sol/gpt-6-astra yet? i need the tracesfrozen21d ago15K147190161.15%
Today, @MidcenturyAI introduces MC-EgoHands, our frontier 3D harness for human egocentric video. PA-MPJPE across four public hand-pose datasets: > ARCTIC - 3.67 mm > HOT3D - 3.98 mm > H2O - 5.24 mm > FreiHAND - 4.30 mm https://t.co/1At5fuDXWFfrozen21d ago10K125790541.91%
if you're a cracked researcher or know someone who is, do have them apply for our research residency program, cohort 2. in our first cohort, we already had an author of NVIDIA's Cosmos 3, a CoRL award winner, and researchers from CMU, UCL, and Brown, with publications at NeurIPS and ICRA.frozen8d ago9.6K1456301052.69%
https://t.co/5it5Nyo7Y4frozen23d ago9.6K40313141284.78%
mom 😭😭 https://t.co/gyFIowXnH6frozen17d ago9.4K13203071.51%
Navratri dump ✨ https://t.co/UyrbWhTCaQ2d ago9.3K24401152.69%
i've wasted some pockets of my own personal time over the last three months that i could've used to improve myself. i have put my blinkers on again. this fresh start is pretty awesome. i'll spend all this new time learning new things and growingggggfrozen8d ago9.1K311230253.75%
don't worry, kids. my prophecy will speak for itself https://t.co/zgMJb9IjGYfrozen27d ago9.0K122330271.72%
first principles thinking at its peakfrozen7d ago8.8K452620225.48%
thanks. i’ve noticed some of it lowkey carries over once i am done with some short bursts of 1-min blitz games. in work and in life, i start seeing the next couple steps without really trying. pattern recognition now comes naturally to me in everything i do, and so my brain neurons fire really quickly when i see a problem and tries to come up with 5 different solutions(classic chess subconscious). i find myself holding a few options in my head during meetings.frozen12d ago8.2K162131372.48%
i have created a control room panel in my cmux software factory setup that updates me realtime for all the flow of work of implementation, review, etc. planning is done by gpt-6-astra-high. the lead orchestrator can switch the models(terra, luna, sol) for any sub agents it spins up based on the complexity of the problem. it also highlights me the current panel i should direct my focus next for any follow up messages if needed.frozen17d ago7.2K1863601014.10%
i’ve spent 4 days at popvax’s office a few months ago and personally saw the talent density, the lab, and the insanely ambitious work the team is pulling off every single day. @sohamsankaran is a genius, and i’m very bullish on popvax and india’s bio scene from india -> world.frozen11d ago6.8K148620322.76%
i am a power of kivi, definitely recommended. go now and installfrozen17d ago6.1K77120201.65%
Happy Navratri everyone 🙌💫 May She shower her divine blessings upon you, guide and protect you, and fill your home with happiness, health, and prosperity. https://t.co/4yOESfhOkYfrozen11d ago5.4K497145089.67%
was part of the initial conversations while this high-throughput wet lab was being planned. happy to see navvye and tyler both pull this off in so little time. two of the most cracked af people i’ve worked with. the era of bio is here. godspeed 🚀frozen13d ago5.1K81212121.91%
“amateurs seek the sun. get eaten. power stays in the shadows”2d ago4.4K16944094.23%
someone at anthropic thought publishing this blog was a good idea 🤦‍♂️ https://t.co/reGLsM1pbkfrozen22d ago4.0K70120152.20%
deepdiving into deepseek-ai/deepinfer this weekend. https://t.co/FERLIwKWtP2d ago4.0K151131214.46%
Generic LLMs fail at serialized fiction for a boring reason: They have no long-horizon reward function.  They optimize for the next sentence.  Pocket FM optimizes for the next 40 minutes of listening. Sherpa’s prose engine was trained on Pocket’s own catalog + drop-off data + a 40-point scene checklist. Memory tracks who knows what, 4 seasons later. Planning works season → arc → episode → scene.  In a nutshell, a world engine sitting on a closed-loop dataset nobody else has.frozen22d ago3.3K37110292.08%
walk away13d ago2.7K3900001.46%
had a fun weekend playing with opensource argus and optimising it to get state-of-the-art 50.2% on semantic F1 on WGO Bench. today, frontier models like gpt-6.1-sol and gpt-6-astra have made it dramatically easier to understand what’s happening in robot videos - visually, temporally, and at the level of individual actions. we did a few optimizations keeping argus as the foundation harness to get these results w gpt-6.1-sol-high!frozen17d ago2.6K4620082.15%
Today, @MidcenturyAI is introducing MC-PhysBench: A Time-to-Failure Physics Benchmark for World Models. The benchmark gives a world model one still frame from the end of a 1s simulated physics clip + a short text desc. The WM must generate the next 2 seconds. MC-PhysBench then asks: how long does that continuation remain inside the range of futures that physics permits? The score is Time To Failure: the first moment a physical violation persists for at least 0.3 seconds. The bench also produces survival curves and categorizes the specific failure type, such as: > an object vanishing > moving incorrectly > passing through another object > or the camera drifting. Each result is averaged over 50 identical episodes per scene, so it reflects a distribution of failures rather than one lucky/unlucky generation. Gen-4.5 ranks #1 the benchmark with a 30% score, Cosmos 3 Super ranks #2 with 21%. Superb work done by @shilpishahh and @B_S_N_Y 👏frozen14d ago2.5K6140093.00%
a research agent hits a much more interesting boundary when the answer doesn't exist in its context. Qualitate can find the right people, collect new observations and bring them back into the research. that's a loop worth building.2d ago2.4K24210151.76%