Mat Keep

Recovering tech addict. Working across product dev and GTM for #data, apps, and #AI. Avid traveller and #Pompey fan

[Mat Keep @matkeep]

https://t.co/6rYqR7kFzj - a great way to assess how open most #AI and #software projects truly are - designed for tired old gits like myself
https://t.co/dJT6fk6vCE

[Mat Keep @matkeep]

RL training is full of silent failures that look like someone picked the wrong hyperparameters. @FireworksAI_HQ found one of the real culprits: training and inference quietly disagreeing on the numbers. Here's the proof, and what to do about it: https://t.co/cxVIzzovaF

[Mat Keep @matkeep]

A master class in diplomacy from @dittycheria

[TBPN @tbpn]

MongoDB interim CEO @dittycheria on Meta poaching CJ Desai:

"CJ made a decision, and I'll let you pass judgment."

"What I will say is I got a note from a good friend who says, 'At the end of the day, no one's going to remember how much money you make. They're going to remember what kind of leader you were and what kind of person you were.'"

"I'll just leave it at that."

[Fireworks @FireworksAI_HQ]

Today we're launching the Specialized Intelligence Index (SII): one destination for real-work benchmarks across industries, built by the teams that use them every day.

Hear from Fireworks co-founder @the_bunny_chen on the importance of specialized benchmarks:

[Fireworks @FireworksAI_HQ]

We’ve just added three new speakers to our incredible Fireworks Forge lineup!

We are proud to welcome @BrendanFoody @pirroh and @jefftangney to the stage as we bring together the people, teams, and companies building their own frontier on open models.

[Mat Keep @matkeep]

Had the privilege of working with the @phylo_bio team on telling the company’s journey to open weights models #agenticAI

[Fireworks @FireworksAI_HQ]

Open models are becoming increasingly capable in biology. Serving them well is harder.

Fireworks gives @phylo_bio day-zero access and a fast serverless path, putting research agents in front of more scientists.

Learn more: https://t.co/vwiWnBFQY4

[Mat Keep @matkeep]

Rented intelligence is the same model everyone else calls. Owned intelligence is trained on what only you know, and it gets better every time you use it.

Join the @FireworksAI_HQ webinar on making the leap to specialized intelligence.

https://t.co/CXVn0OymDt

[Mat Keep @matkeep]

Training API is now generally available.
Bring your own loop, @FireworksAI_HQ runs the infra.
3 ways to train
→ Training API (bring your loop)
→ Managed Training (bring your data)
→ Fireworks Lab (embedded #ML researchers)
Train past the frontier: https://t.co/IDTu2kvxlt

[Mat Keep @matkeep]

Nearly 2x more legal agent tasks completed. Same cost per task. @harvey and @FireworksAI_HQ post-trained #Kimi K3 into Tenet using async RL. Summary of how specialized intelligence for #legal workloads stacks up https://t.co/Uz7fA07BSs

[Mat Keep @matkeep]

The team that developed @FireworksAI_HQ Nexus have done an amazing job at eliminating #softwareengineering dependency on closed #AI models. Instead Nexus brings the most efficient model to the task, along with full cost transparency. Here's an overview https://t.co/uTBtALTfbA

[Mat Keep @matkeep]

@IanGrayu6ur @JeremyClarkson Bot

[Mat Keep @matkeep]

Voyage AI models now run natively on Fireworks, the first and only dedicated inference platform @VoyageAI by @MongoDB has partnered with.
Embed, retrieve, rerank, generate: your full retrieval pipeline on one platform, next to the open models you own.
https://t.co/NpZ3FxNtAy

[Mat Keep @matkeep]

Available today - #kimiK3 now available on @github #copilot, served on @FireworksAI_HQ https://t.co/BUotkpTXfh

[Mat Keep @matkeep]

Kimi K3 on @FireworksAI_HQ : matches #Opus 5 at up to 5x lower cost per task, #1 open model in frontend code. Serve day-0, then fine-tune from your laptop. Start with the open model as your default, route to closed only where the task demands it: https://t.co/qcnO1oY9ln

[Mat Keep @matkeep]

Benchmarking #kimiK3 and #fable on @FireworksAI_HQ - results across over 1k tests show the advance of open #AI models https://t.co/vDHMK80cpY

[Jesse Zhang @thejessezhang]

https://t.co/N2gSVInZVT

[Mat Keep @matkeep]

As #AI models become commodities, the competitive advantage shifts to the #harness. Great read on why owning your execution loop beats vendor lock-in every time.
https://t.co/nmAVesp0vV

[Aparna Dhinakaran @aparnadhinak]

https://t.co/j4beh1fjcR

[Mat Keep @matkeep]

The #AI debate often asks: "How many jobs will AI replace?". This research suggests a better question: "Which companies will use AI to create more growth?"

The biggest AI adopters were also the fastest hirers. AI appears to reward ambition, not cuts https://t.co/PwVzHis1mU

[Mat Keep @matkeep]

Everyone's chasing better models. The bigger competitive advantage is better data pipelines.

The teams that build a data flywheel that continously feeds quality data to models will make their #AI smarter ahead of their competitors @ApacheAirflow https://t.co/foCDL1648X

[Mat Keep @matkeep]

We've spent two years obsessing over #GPUs, #AI model benchmarks and token pricing. The next question is much harder: how much intelligence are we actually buying?

The companies that answer that will define the next phase of AI:
https://t.co/NoS4yLAXFv