Matija Čupić

558 posts

Matija Čupić banner
Matija Čupić

Matija Čupić

@matteeyah

MSc Information Technology | Former Founder | Previously: @Pennylane_tech, @GitLab

Belgrade, Serbia Katılım Mayıs 2012
124 Takip Edilen199 Takipçiler
Matija Čupić
Matija Čupić@matteeyah·
@garrytan What’s the ballpark estimate of how many tokens are you using per week / month? I don’t want to try GBrain because I might like it but not be able to afford the tokens 😭
English
0
0
1
26
Matija Čupić retweetledi
Y Combinator
Y Combinator@ycombinator·
We’re bringing Jensen Huang to Startup School for a fireside chat with @garrytan! From co-founding @nvidia in 1993 to building the backbone of the AI era, Jensen helped turn GPUs into the engine of modern computing. Apply to attend: events.ycombinator.com/startup-school…
English
37
65
789
89.2K
Matija Čupić
Matija Čupić@matteeyah·
@ritwikpavan What benefit does this have over existing FotoFinder systems? I’m not familiar with how Swan works, but it seems more expensive, more complex, and less comprehensive than current gen FotoFinder systems. 🤔
English
0
0
1
122
Ritwik Pavan
Ritwik Pavan@ritwikpavan·
JUST IN: Skin exams are getting automated. SquareMind just raised $18M to build a robotic system that scans your entire body and tracks every mole over time. • Swan robot captures full-body dermoscopic images in minutes • Tracks new and changing spots across visits • Replaces spot-check exams with total skin coverage • Creates a time-series record for earlier melanoma detection • Plugs directly into dermatology clinics Robotics is going to reshape healthcare.
English
148
448
5K
1.7M
Matija Čupić
Matija Čupić@matteeyah·
@forgebitz How did you end up deciding to use Cursor as your go to AI agent harness?
English
0
0
1
27
Klaas
Klaas@forgebitz·
promptwatch devs are the top 1% cursor users apparently (we are hiring)
Klaas tweet media
English
15
1
52
5K
Matija Čupić
Matija Čupić@matteeyah·
@andruyeung Yes! Lovable's not even close to making their product seamless for non-technical users. This is a huge gap in the market. Bring AI to the masses!
English
0
0
1
78
Andrew Yeung
Andrew Yeung@andruyeung·
Only 0.3% of the world pays for an AI service. There is a massive opportunity to build AI for ‘normies’ — the everyday person who is not plugged into Silicon Valley but can still benefit from what AI has to offer. And I think in-person experiences are the best way to do it.
Andrew Yeung tweet media
English
30
6
71
8.6K
Matija Čupić
Matija Čupić@matteeyah·
@garrytan Food production is one of the most undisrupted markets, IMHO. The disruptions that did happen, all had a net negative effect on the health of humans…
English
0
0
1
11
Garry Tan
Garry Tan@garrytan·
Everybody eats. Improving our food supply and safer more efficient agriculture is one of tue most direct ways for AI to help people
Calley Means@calleymeans

@ycombinator @garrytan Thank you for the leadership @garrytan in spurring leadership in technology for our farmers - who are the key to reversing our skyrocketing rates of chronic disease.

English
34
17
277
33.7K
Matija Čupić
Matija Čupić@matteeyah·
@forgebitz See, I thought the same thing about code review. AI review was mostly a gimmick. Until Claude launched Ultrareview / hosted GitHub reviews. It’s, on average, nuanced as much as a human review is. It’s only a matter of time…
English
0
0
1
38
Klaas
Klaas@forgebitz·
custom support is solved by ai until you have to speak to ai customer support and realize is completely useless and a very hard problem to solve
English
53
7
269
8.7K
Mow
Mow@Mow_exe·
@1realFormula Bro thinks American cheese is healthy
Mow tweet media
English
28
20
6.1K
373.4K
Formula🌵
Formula🌵@1realFormula·
What’s this logic 😭?
Formula🌵 tweet media
English
3.2K
5.2K
87.1K
27.7M
Bryan Johnson
Bryan Johnson@bryan_johnson·
I got put under and had probes inspect my gastrointestinal tract. Down my throat and up the anus. It was my first bidirectional endoscopy. Right before the anesthesiologist injected, I thought "This is the end. This is how I die. What could make the internet happier?" Then it was lights out. This was important to do because colon cancer is now the #1 cancer killer under 50, and rising fast. Early onset colorectal cancer incidence has more than doubled since 1994, climbing roughly by 3% per year in 20 to 49 year olds and 8% in 20 to 29 year olds. Even though I routinely do all sorts of painful things to my body and mind for this project, this procedure had been weighing on me. I didn't want to go under and, you know, didn't love the idea of the probes being snaked through my body. Glad it's over and it honestly was not as bad as I had anticipated. Here are my results: + doctor gave me a 10/10 score + no polyps + no inflammatory bowel disease + no diverticulosis, a common colon-aging marker + we are awaiting biopsy results Colonoscopy screening can save your life. A meta-analysis of over 4.7 million people found colonoscopy was associated with 52% and 62% reduction in colorectal cancer incidence and mortality. A regular full body MRI (I get one every 6 months) can't reliably detect early colon lesions or polyps. A colonoscopy remains the gold standard for both detection and removal. The recommended screening age has been lowered from 50 to 45 in 2021. Half of early colorectal cancer cases now fall in 45 to 49 year olds. Obesity is a genuine colorectal cancer risk factor. A meta-analysis involving over 66,000 participants found obesity raises early colorectal cancer odds by roughly 50%. Yet it is unlikely to fully explain the under 50 surge. A new study surfaced an unexpected culprit. Across 10 cohorts and 29 lifestyle and environmental signatures, comparing tumor DNA methylation in early-onset (<50) vs late-onset (≥70) colorectal cancer, one signal stood out: the herbicide picloram. Early onset tumors showed ~3 fold higher odds of carrying the picloram methylation signature in the discovery cohort, and 1.77-fold across all 10 pooled cohorts (114 early onset vs 372 late onset). Across 94 US counties over 21 years (1992 to 2012), picloram-use intensity correlated with increase in EOCRC incidence, the most robust signal among 62 pesticides tested. Early onset tumors carried a lower obesity methylation signature than late onset, suggesting that environmental toxins, more than metabolic dysfunction, are the dominant epigenetic driver in young patients. Epigenetic drift drives biological aging and most age-related disease: chemicals assumed safe because they aren't directly genotoxic may still predispose us to cancer and chronic disease over decades through methylation and gene-expression disruption It's time to ring the alarm: every additive chemical in our food, water, and environment needs re-evaluation through a long term, population-based epigenetic and gene expression lens, not just acute genotoxicity assays Epigenetic disruption by environmental toxins is likely a key driver.
Bryan Johnson tweet media
English
167
83
1.3K
292.3K
Matija Čupić retweetledi
Elon Musk
Elon Musk@elonmusk·
Some pics from when Jensen delivered the first @Nvidia AI system to @OpenAI
Elon Musk tweet mediaElon Musk tweet mediaElon Musk tweet media
English
851
1.6K
15.9K
5.4M
Matija Čupić
Matija Čupić@matteeyah·
@brightafia That’s cool. Totally viable. But how do you get to 180g-190g of protein, which is a more realistic target for most men that do weights? 2-2.2g / kg bodyweight. That’s almost 8 eggs and 400g of yoghurt a day. Not very viable.
English
3
1
9
1.7K
Bright.web3
Bright.web3@brightafia·
HOW TO HIT 100g PROTEIN WITHOUT SUPPLEMENTS • 4 eggs — 24g • 150g chicken — 45g • 1 cup beans — 15g • 200g yogurt — 20g No powders. No excuses
English
59
432
4.9K
189.4K
Matija Čupić
Matija Čupić@matteeyah·
@shawmakesmagic The impression that I have is that YC mostly judges the people on an application, not the idea itself. Ideas change, most startups pivot, but you still need a rock star team to pull through all of that.
English
0
0
3
1.5K
Shaw (spirit/acc)
Shaw (spirit/acc)@shawmakesmagic·
Ironically I applied to YCo once with this idea and they rejected me I would never apply again, just giving Garry your ideas for free so he can vibe code and claim credit out of ignorance for an industry he himself gatekeeps
Garry Tan@garrytan

For GBrain I built a proper eval harness. 145 queries, Opus-generated corpus. The retrieval stack uses graph based, vector based and Grep based strategies in combination. The graph layer is worth +31 points on precision. Vector-only misses 170/261 correct answers that the full system finds. Keyword + vector + graph are three separable wins, each load-bearing. Standard information retrieval metrics: the same ones Google uses to measure search quality. Precision at 5: You ask a question, the system returns 5 results. How many of those 5 are actually useful? If 3 out of 5 are relevant, P@5 = 60%. It measures: am I wasting your time with junk results? Recall at 5: For a given question, there might be 3 pages in the entire brain that are genuinely relevant. If the system finds all 3 in its top 5, R@5 = 100%. If it only finds 1, R@5 = 33%. It measures: am I missing things you need? High precision = low noise. High recall = nothing slips through. GBrain's 97.9% R@5 means it almost never misses the right answer. The 49.1% P@5 means about half the results are relevant — which is good when you realize that for most queries there are only 1-2 right answers out of 17,888 pages, so 2.5 hits out of 5 is strong signal. Entity resolution is zero-LLM-call: regex extracts typed links (works_at, invested_in, founded) on every write. Re-embed on write not on a timer, so decay = stale pages, and stale pages get rewritten when new info lands. Scorecards: github.com/garrytan/gbrai…

English
45
23
950
162.1K
Michael
Michael@michael_chomsky·
Ok how tf do I review this PR?
Michael tweet media
English
185
10
516
86.9K
Hubert Thieblot
Hubert Thieblot@hthieblot·
The longer I'm on X, the more I realize: Founders with small accounts are the most interesting ones. - too busy building to posture - 0 Ego, they just wanna win - keep posting with 0 likes - the world isn’t rooting for them yet but I will Tell me what you are building
English
908
69
2.7K
88.1K
Matija Čupić
Matija Čupić@matteeyah·
@bstaples @sytses I _love_ GitLab. Use it daily. But it’s sooo behind the curve on AI, it’s not even funny. They should just drop most of what they’re working on and focus exclusively on making AI their bread & butter offering.
English
2
0
3
436
Bill Staples
Bill Staples@bstaples·
For those teams looking for alternatives to the GitHub meltdown, check out this recent customer chat I had with the software team behind Rivian/Volkswagen. They run GitLab dedicated, which means they get all the benefits of SaaS, but with dedicated infrastructure and isolation at enormous scale. One of the keys to our success: you are always in control of how and where you run GitLab. GitLab serves more than half of the fortune 100, and hundreds of thousands of other organizations around the world, why not yours? x.com/bstaples/statu…
Tom Elliott@theotherelliott

This GitHub incident is insane. Merge queue commits have been reverting previously merged commits at random. This not only breaks the mental contract teams have with Git in general, but is subtle enough to be really hard to unravel after the fact. githubstatus.com/incidents/zsg1…

English
4
7
76
18.2K
Matija Čupić
Matija Čupić@matteeyah·
@Txp_RBI_Xctuxl @paulg It’s definitely trending the same way BMW is, but it still has flagship models with the classic Mercedes look & feel
Matija Čupić tweet media
English
0
0
1
22
Matija Čupić
Matija Čupić@matteeyah·
@dr_ericberg Nutrition and cholesterol. It’s the number one cause of death worldwide. It’s not that it’s misunderstood - but the whole world is being on eating wrong for some reason.
English
0
0
2
53
Dr. Eric Berg DC
Dr. Eric Berg DC@dr_ericberg·
Which health issue do you think is misunderstood the most?
English
110
4
58
11.6K
Leo
Leo@Leooweb3·
If you had a $5M networth. Would you spend $250k on a Porsche 911?
Leo tweet media
English
3.2K
327
8K
560.8K
Matija Čupić
Matija Čupić@matteeyah·
@LxngevityLab Except not eating from evening to morning doesn’t actually trigger autophagy. It takes at least 24h of fasting to notice autophagy. After 48h it should be more pronounced.
English
0
0
1
30
LongevityLab
LongevityLab@LxngevityLab·
🚨 Eating 3 meals a day is accelerating your aging. David Sinclair exposed it:
English
3
14
51
5.7K