Chris Klaus

CEO of Fusen. Connecting students with mentors, investors, and funding opportunities through our Fusen accelerators. @cklaus.bsky.social on Bluesky.

[Chris Klaus @cklaus1]

@kaustabpal @blackwellboy When you run a bunch of subagents in parallel and building multiple projects in a loop, you'll want C=32.

[Cohere @cohere]

We encourage developers to share their builds with us and give feedback to shape future iterations. Let’s shape the future of sovereign AI together.

Download: https://t.co/2C3eqGF95i

[Chris Klaus @cklaus1]

Testing Mythos...

[Chris Klaus @cklaus1]

LFM model beating Gemma + Qwen for the 1B - 2B SLM.

[Chris Klaus @cklaus1]

@ycombinator Congratulations on the launch! Big opportunity to go after!

[Aidan Pratt 🥑 @aidan__pratt]

@ycombinator Want to try us out? Book a demo here - https://t.co/f3PwlYpZt8

[Y Combinator @ycombinator]

Autostep finds the repetitive work hiding inside your company, then helps you fix it. It context mines how your team works across emails, decks and reports. Then it proactively spawns agents on that context so the work never gets repeated again.

Congrats on the launch, @aidan__pratt!

https://t.co/GtjRQsQjvP

[Chris Klaus @cklaus1]

Getting Gemma 4 26B on my 5090 GPU required some patches, giving back.

Hope the patches works for others.

[Chris Klaus @cklaus1]

meta-ads-cli now supports Threads ads, WhatsApp Status ads, Advantage+ campaigns, full placement targeting across 6 platforms.

All from the command line. No Ads Manager needed.

I did this mostly for my superagent for ads mgt.

https://t.co/LQlhjeI8FZ

[Chris Klaus @cklaus1]

All three vanilla Qwen3.5 NVFP4 quantizations fail identically at tool calling. The Claude Opus reasoning distillation is what makes tool calling work. That model was specifically trained/distilled for agentic capabilities, and it shows.

Has a big impact on your local Agents, like OpenClaw.

[Garry Tan @garrytan]

Just added @greptile to my @garryslist repo and... wow how did Iive without you Greptile?

[Sean Henry @seanhenry]

Disruption isn’t new, but it 𝘪𝘴 constant.

That’s why Stord continues to scale a tech-enabled, global fulfillment network, including our recent Shipwire acquisition, to help independent brands compete with Prime-level expectations even as conditions change.
Resilience is built.

Appreciate Bloomberg TV for having me on to talk about what e-commerce brands are navigating right now.

[Sean Henry @seanhenry]

Excited to share that 𝗦𝘁𝗼𝗿𝗱 𝗵𝗮𝘀 𝗮𝗰𝗾𝘂𝗶𝗿𝗲𝗱 𝗦𝗵𝗶𝗽𝘄𝗶𝗿𝗲 𝗳𝗿𝗼𝗺 𝗖𝗘𝗩𝗔 𝗟𝗼𝗴𝗶𝘀𝘁𝗶𝗰𝘀!

This adds 12 locations, strengthens our EU/UK presence, and expands access to CEVA’s global warehouse network through existing Shipwire agreements.

It’s another step forward in building the infrastructure and technology brands need to deliver better consumer experiences at scale.

We are honored to welcome the Shipwire team and customers to Stord.

You can read more here → https://t.co/TnJzuNF9jF

[Daksh Gupta @dakshgup]

animations are a great way to explain things quickly

[Francois Chaubard @FrancoisChauba1]

Last night, @agupta and I hosted a great dinner with 14 professors at #NeurIPS2025 from leading academic labs across the US, and many cited compute in academia as "abhorrent". Out of curiosity I just pulled these stats. This is insane. To do meaningful AI research today you need at least 1 GPU/student. Likely 8+ to be honest. The best university (Princeton) is at 0.8 GPUs/student. Stanford is at 0.14 GPUs/student. Marlowe (Stanford's "super cluster") has only 248 H100s for the whole CS Dept to use. Every frontier lab has >100k.

This needs to be fixed.

[agrim singh @ ai engineer world fair @agrimsingh]

single greatest swag item i’ve received at a tech event. @greptile @dakshgup nothing like ballmer screaming “developers, developers, developers”on demand (and the matcha is a good touch too!) love it

@aiDotEngineer

[Tech Square ATL @TechSqATL]

Born in Tech Square and founded by Georgia Tech alumni Pavleen Thukral, Magic Data is building an AI “virtual data analyst” that automates the cleaning, organizing, and analysis of messy enterprise data. Instead of hours of prep, teams can ask questions in plain English and get answers in minutes.

🚀 Founded by a Georgia Tech alum and HackGT co-founder.
📊 Built to help mid-tier and large enterprises save millions.
📍 Active around Tech Square’s Biltmore Innovation Center.
💪 Backed by entrepreneurs like @cklaus1

Read the full story 👉 https://t.co/jsJl1JZuQU

[Baxate @Baxate]

Now anyone can pre-train your own model in 4 hours. Incredible work by @karpathy to open source and democratize some of the most important educational resources in the AI Era. The “Eurekas Per Second” in nanochat is something you must experience.

The NVIDIA Brev team has created a launchable (see below) for you to try!

Simply click “Deploy Launchable” and a GPU will be provisioned in the cloud, and nanochat will begin training!

The first 10 developers to deploy will be able to train their own GPT 2 style model completely for free. Happy hacking!

[Chris Klaus @cklaus1]

@AngelList Im surprised it's not a higher %.