Jacques Gariepy @JacquesGariepy

Software engineer - https://t.co/d6ia3tWB8d Québec, Québec, Canada Joined May 2014

Tweets

682
Followers

152
Following

1K
Likes

709

Logan Kilpatrick @OfficialLoganK

12 hours ago

google.com/ai

187 70 1K 196K 201

Jacques Gariepy @JacquesGariepy

12 hours ago

By 2027, inference runs on distillation, quantization, and compiler-aware training. Use int4 with outlier paths, selective FP8, fused kernels. Tile, cache, prune attention for long context. Winners treat inference as systems work. Quality-aware compilers are the edge.

0 0 0 10 0

Jacques Gariepy @JacquesGariepy

22 hours ago

Emergence comes from constraints, not complexity. Sharp interfaces let small agents scale smart. Write laws before policies. Example: no spend without signed intent + matching state-hash. Ship constraints first; behavior emerges.

0 0 0 8 0

Jacques Gariepy @JacquesGariepy

a day ago

We scaled a DSpy swarm, it broke. Reward hacking fooled our metrics. Invariants and a budgeted planner fixed it. Throughput dipped, reliability soared. Constrain first, then optimize.

0 0 0 26 0

Jacques Gariepy @JacquesGariepy

2 days ago

Pattern: Guardian-Agent for autonomous safety. Worker proposes; guardian enforces rules, learned critics, and state-diff checks. Deny-by-default, log readable diffs, and budget every exception.

0 0 0 4 0

Jacques Gariepy @JacquesGariepy

2 days ago

perplexity.ai/join/p/paypal-…

Ahmad @TheAhmadOsman

2 days ago

perplexity.ai/join/p/paypal-…

98 75 2K 278K 2K

0 0 0 84 1

Lior⚡ @LiorOnAI

3 days ago

You can now run 100B parameter models on your local CPU without GPUs. Microsoft finally open-sourced their 1-bit LLM inference framework called bitnet.cpp: > 6.17x faster inference > 82.2% less energy on CPUs > Supports Llama3, Falcon3, and BitNet models

75 435 3K 340K 3K

Download Video

Tenstorrent @tenstorrent

4 days ago

BYO Supercluster by @0xCodyS on @huggingface huggingface.co/blog/codys12/d…

3 5 25 2K 12

Prime Intellect @PrimeIntellect

a week ago

Introducing the Environments Hub RL environments are the key bottleneck to the next wave of AI progress, but big labs are locking them down We built a community platform for crowdsourcing open environments, so anyone can contribute to open-source AGI

117 398 3K 1.2M 2K

Download Video

Jacques Gariepy @JacquesGariepy

a week ago

With Intel® Arrow Lake Z890 ATX. What do you think and I'm going to be able to ride what do you think LLM is?

2 0 2 331 0

Download Image

Ahmad @TheAhmadOsman

2 weeks ago

IMPORTANT: like and reply to this tweet (to avoid bots) before requesting to join; it shows you read the instructions introducing our X community for local llms, self-hosting, and hardware requests without answers and without like+reply = denied

469 35 719 39K 109

Min Choi @minchoi

4 weeks ago

ChatGPT Reddit is blowing up ChatGPT users are literally cancelling over GPT-5

211 137 1K 227K 355

Download Image

The Lunduke Journal @LundukeJournal

4 weeks ago

Today, Linus Torvalds told a Google engineer that his code (updating RISC-V support in the Linux kernel) is “garbage” which “makes the world actively a worse place to live”. Adding that the Google engineer’s code needs to “get bent”. As you might have guessed, Torvalds has…