Hasan

[Hasan @hasantkose]

my vocabulary shifted:

from: xfn, experiment, psc, stat-sig, be
to: wedge, moat, market-fit, tam, distribution

[Hasan @hasantkose]

@ycombinator just dropped their Fall RFS, one of them is AI-native compliance infrastructure last week at the hackathon we built similar thing: policy fuzzing, monitoring regulatory changes, flagging anomalies, and alerting you when something breaks. guess we spotted a real pain

[Hasan @hasantkose]

hacked at the @Stanford @GoogleDeepMind hackathon this weekend, 3 hours to code, a pitch video, and a demo. We built ClauseCrash: CI for policies, watch upstream regulations, auto re-test everything downstream when they change.

[Hasan @hasantkose]

I can see the direction the industry is going. Perfect today for anything demo-shaped; dashboards, simple sites, internal tools. Not yet for anything spec-shaped. Either way, a great experience, the idea-to-working-prototype gap has never been this small.

[Hasan @hasantkose]

hacked at the @Stanford @GoogleDeepMind hackathon this weekend, 3 hours to code, a pitch video, and a demo. We built ClauseCrash: CI for policies, watch upstream regulations, auto re-test everything downstream when they change.

[Hasan @hasantkose]

cons: making it match an actual spec takes forever. We even wrote a planning spec first, still slow to converge. Impressive first-glance demos; the last 30-40% of real functionality is still on you, and ofc it is harder than first part.

[Hasan @hasantkose]

More than anything, I’ll remember the people — colleagues who made hard problems easier, and managers and mentors who took the time to invest in me and push me forward. I’m deeply grateful to all of them.

Excited for what’s next.

[Hasan @hasantkose]

Today is my last day at Meta, after almost 7 years.

I joined in 2019, just a year after moving to the U.S., and Meta became more than a job — it’s where I built my life in a new country.

[Hasan @hasantkose]

I got to work on products used by billions of people, alongside incredibly kind, talented people who were always willing to help. I grew more than I ever expected, both personally and professionally.

[Hasan @hasantkose]

@analysmoo Good stuff, do you anticipate only pro videos with wide and high angle for analyses ? I am testing (also exploring similar ideas) with Xbotgo or Veo videos, the models are having issue with tracking as often confuses w bad angle and can't detect players.

[Hasan @hasantkose]

@peacock can’t watch France vs Norway game #worldcup

[Hasan @hasantkose]

Sprint = centroid velocity sustained for 45 frames. All Swift on-device w CoreML dtype handling on the Neural Engine

[Hasan @hasantkose]

Exploring EdgeTAM + YOLOv8 on iPhone — tap a player in a soccer video, tracks them and auto-detects touches, dribbles, kicks and sprints. All on Neural Engine ⚽

[Hasan @hasantkose]

YOLOv8n(@ultralytics ) detects the ball every 4 frames. Bottom of the player mask = feet. Ball near feet = touch. Stays there = dribble. Ball disappears after contact = kick.

[Hasan @hasantkose]

@skalskip92 how does it work for side camera? for example if a parent records their kids playing using iPhone, would this model be able to track players from side view or bird view is mandatory?

[Hasan @hasantkose]

@geniegetsme chats are getting slower when the conversation grows, after 10+ messages, I can’t even scroll the thread and it freezes. Any upcoming build to improve that?

[Hasan @hasantkose]

@AodenTeoMT Awesome, I played with live preview, and it is pretty good. Tho, it started hallucinating at the end of custom text. Happy to share screen recording via dm. Overall it is impressive.

[Hasan @hasantkose]

Just launched Pitch Pulse — built it for soccer players who wear Apple Watch. Heat maps, sprints, bursts after every game. Check it out 👇

[Pitch Pulse @pitchpulsetrack]

Pitch Pulse Tracker is live. 🚀

GPS heatmaps. Sprint detection. Heart rate zones. Training load. All from your Apple Watch after every match.

Built for soccer players who want to understand their game. Link in bio 🙌