Securade.ai

Generative AI based Video Analytics OSS |
Winner of AI for Public Good @startupmhakumbh |
HPC AI Innovation Challenge ( |
Safety Startup (

[Securade.ai @securadeai]

AI can detect safety violations faster than your site supervisor.
Here’s how we're doing it — with open source.
🧵👇

[Securade.ai @securadeai]

@ConstructionEqt @DEVELONna Looks great, and if you are looking for an open-source solution that can do video analytics try HUB - https://t.co/K1J9zdUVBy

[Securade.ai @securadeai]

Clear video evidence is critical. That's why we built Securade Hub an open-source edge video analytics for high-quality, intelligent surveillance.

🔗 https://t.co/K1J9zdUVBy
📰 https://t.co/aKiHm1SSXh

(2/2)

[Securade.ai @securadeai]

A recent Singapore condo incident where a resident suffered fatal injuries after being startled by a dog highlights why CCTV quality matters for safety AND liability.

The judge noted the footage was "poor quality" at "one frame per second" and couldn't capture the detailed sequence of motion.

(1/2)

[Securade.ai @securadeai]

@AndrewYNg Any details on how the model was trained or if the base model itself is open-weight?

[Securade.ai @securadeai]

@JayantBhandari5 What is decadent in this? Have you never been to an Indian wedding?

[Securade.ai @securadeai]

The most productive thing I did this week was absolutely nothing.

15 minutes of silence every morning. No phone, no agenda, just stillness.

Last week, during one of these pauses, I caught a $50K mistake we'd been making for months.

Speed without direction is just expensive motion.

The best decisions happen in the space between stimulus and response.

What could you solve by slowing down?

[Securade.ai @securadeai]

🚨 Just found something that could save your team 90% on text classification costs

Someone built 17 production-ready classifiers that run locally and learn continuously:

📊 93% avg accuracy with just 100 examples per class ⚡ 90ms inference (no API calls needed)
💰 $100/month vs $600/month for OpenAI
🔄 Add new classes without retraining

Perfect for:
• Email routing & security
• Support ticket classification
• Document processing
• Fraud detection
• Content moderation

The adaptive learning part is genius - they actually get BETTER over time as you use them.

All models available on HuggingFace: https://t.co/y0qISOBivg

Full breakdown: https://t.co/MQQZ3XPIgp

This is exactly what we needed for our document pipeline. No more burning API credits on simple classification tasks 🔥

#MachineLearning #AI #TextClassification #LocalLLM

[Securade.ai @securadeai]

@OfficialLoganK @_mohansolo How did he fumble the 3B offer from @OpenAI ?

[Securade.ai @securadeai]

@tenderizzation Upgrade it on a @nvidia Jetson and experience the real trauma. That's exactly why we have a full wiki devoted on it here - https://t.co/6xvKbWb1vj

[Securade.ai @securadeai]

@AravSrinivas @PerplexityComet Are you serious? Why would they use a subpar search engine as default?

[Securade.ai @securadeai]

@soumilrathi @seatedro It is funded by YC and is following the playbook as PearAI did .

[Securade.ai @securadeai]

@Inveeest @soham_btw @garrytan @ycombinator @trypearai @trypearai was also not a fork they had originally copied the code and removed attribution exactly like what was done here.

[Securade.ai @securadeai]

@that_anokha_boy @soham_btw @garrytan @ycombinator Yeah it is the @trypearai playbook ...

[Securade.ai @securadeai]

@devahaz I believe this is called the @trypearai playbook at @ycombinator …

[Securade.ai @securadeai]

It is 5 kg per person per month. It is what is given for free under PDS, they are not surviving off it. They are doing quite well, in fact many actually sell the extra rations given to them. Current poverty rate in India is less than 5% see https://t.co/8NTWrnPwip when defined as < 3 USD per day which is ~250 INR per day.

[Securade.ai @securadeai]

Someone should try using openevolve https://t.co/kbzXP8xbS4 with this benchmark.

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements

"To evaluate the ability of AI agents to reproduce results in an active research area, we introduce the Automated LLM Speedrunning Benchmark, leveraging the research community’s contributions on the NanoGPT speedrun, a competition to train a GPT-2 model in the shortest time."

"recent reasoning LLMs combined with SoTA scaffolds struggle to reimplement already-known innovations in our benchmark, even when given detailed hints"

[Securade.ai @securadeai]

@HighyieldHarry It is 2025, he is at @a16z investing in @cluely!