gatienboquet

5 ms·

31.
by gatienboquet 2y ago
Stripe Driven Development : Build a landing page with a stripe payment. If people pay, build the solution. Rince and Repeat for unlimited money.
32.
by gatienboquet 2y ago
Great book on this subject https://www.momtestbook.com/
33.
by gatienboquet 2y ago
> We’re making the same mistake with money that we once made with abstinence-only sex education: pretending that pure knowledge will triumph over emotion, peer pressure, and real-world temptations. Spoiler alert: it won’t. Funny things,
34.
by gatienboquet 2y ago
Added to my website iwontpayforthis.com
35.
by gatienboquet 2y ago
My 8cores CPU goes to 99% when i visit your website. There's something wrong.
36.
by gatienboquet 2y ago
It's me or it's very slow on edge ?
37.
by gatienboquet 2y ago
In a distributed system, the overall performance and scalability are often constrained by the slowest component. This Distributed Llama is over Ethernet..
38.
by gatienboquet 2y ago
Location: Montreal, Canada Remote: Yes Willing to relocate: No Technologies: Containerization: Docker, Kubernetes (k8s, EKS) Infrastructure as Code: Terraform, CloudFormation Serverless: AWS Lambda, Fargate Databases: PostgreSQL, MySQL, Pro
39.
by gatienboquet 2y ago
https://tinygrad.org/#tinybox
40.
by gatienboquet 2y ago
According to NVIDIA. > a single server with eight H200 GPUs connected using NVLink and NVLink Switch can run the full, 671-billion-parameter DeepSeek-R1 model at up to 3,872 tokens per second. You can rent a single H200 for 3$/hour.
41.
by gatienboquet 2y ago
https://gitingest.com/
42.
by gatienboquet 2y ago
Gemini-Exp-1206 got 2M context token, what's preventing Google from fine-tuning it with internal google PR etc ? So in the end you'll get a fine-tuned google llm dev and one finetuned for review, etc etc