Ryan Schellack

Product Marketing @Salesforce. Tweets are my own.

[Salesforce @salesforce]

We signed a definitive agreement to acquire @fin_ai, a customer agent platform providing autonomous, end-to-end AI service agents trusted by more than 30,000 companies globally.

The acquisition will complement Agentforce, providing powerful service agent capabilities across all channels and helping companies deploy fast-to-value, trusted AI agents at scale.

Learn more: https://t.co/bd72ZVAUxn

[Thariq @trq212]

I did a 2 hour workshop on the Claude Agent SDK at AI engineer!

We are still so early to agents, I hope this is useful if you’re thinking of making one.

[Ara Kharazian @arakharazian]

No one is talking about this, but this is the bull case for AI and economic growth.

AI spend is up 60x in construction. 35x in manufacturing on @tryramp.

Faster planning and smarter automation will transform these strategic industries. This is what drives US economic + labor productivity growth.

[Jim Fan @DrJimFan]

I've been a bit quiet on X recently. The past year has been a transformational experience. Grok-4 and Kimi K2 are awesome, but the world of robotics is a wondrous wild west. It feels like NLP in 2018 when GPT-1 was published, along with BERT and a thousand other flowers that bloomed. No one knew which one would eventually become ChatGPT. Debates were heated. Entropy was sky high. Ideas were insanely fun.

I believe the GPT-1 of robotics is already somewhere on Arxiv, but we don't know exactly which one. Could be world models, RL, learning from human video, sim2real, real2sim, etc. etc, or any combo of them. Debates are heated. Entropy is sky high. Ideas are insanely fun, instead of squeezing the last few % on AIME & GPQA.

The nature of robotics also greatly complicates the design space. Unlike the clean world of bits for LLMs (text strings), we roboticists have to deal with the messy world of atoms. After all, there's a lump of software-defined metal in the loop. LLM normies may find it hard to believe, but so far roboticists still can't agree on a benchmark! Different robots have different capability envelopes - some are better at acrobatics while others at object manipulation. Some are meant for industrial use while others are for household tasks. Cross-embodiment isn't just a research novelty, but an essential feature for a universal robot brain.

I've talked to dozens of C-suite leads from various robot companies, old and new. Some sell the whole body. Some sell body parts such as dexterous hands. Many more others sell the shovels to manufacture new bodies, create simulations, or collect massive troves of data. The business idea space is as wild as research itself. It's a new gold rush, the likes of which we haven't seen since the 2022 ChatGPT wave.

The best time to enter is when non-consensus peaks. We're still at the start of a loss curve - there're strong signs of life, but far, far away from convergence. Every gradient step takes us into the unknown. But one thing I do know for sure - there's no AGI without touching, feeling, and being embodied in the messy world.

On a more personal note - running a research lab comes with a whole new level of responsibility. Giving updates directly to the CEO of a $4T company is, to put it mildly, both thrilling and all-consuming of my attention weights. Gone are the days when I could stay on top of and dive deep into every AI news.

I’ll try to carve out time to share more of my journey.

[pash @pashmerepat]

Boris from the Claude Code team explains why they ditched RAG for agentic discovery.

"It outperformed everything. By a lot"

[OpenAI Developers @OpenAIDevs]

MCP 🤝 OpenAI Agents SDK

You can now connect your Model Context Protocol servers to Agents: https://t.co/6jvLt10Qh7

We’re also working on MCP support for the OpenAI API and ChatGPT desktop app—we’ll share some more news in the coming months.

[Jared Friedman @snowmaker]

Amazing comment by @karpathy on the HN story about Operator, posted by Operator.

[Jaya Gupta @JayaGup10]

"Service as Software" is Silicon Valley's hottest buzzword right now.

Everyone's talking about SaaS becoming service providers, but no one's explaining HOW. The answer? After 6 months of research and 100s of startup conversations, we have the answer: Systems of Agents.

We're looking at a $4.6T opportunity.

[Ch. Szandor Knapp @ch_sz_knapp]

All y'all did a great job but my heroine of the day is @heyitsnanya for her tour through the tooling updates for @SalesforceDevs. Great stage presence and rhetoric.

[Elad Gil @eladgil]

From @davidtsong on my team

Cost of 1M tokens has dropped from $180 to $0.75 in ~18 months (240X cheaper!)

[Ray Fernando @RayFernando1337]

Building app with Cursor, Claude Sonnet 3.5 code stream https://t.co/8RIhHlnYoG

[Riley Goodside @goodside]

This feature really does change things. Gemini had cache, but only for costs — Claude claims 90% less cost *and* 85% less latency for cached tokens.

Even if it can’t replace all SFT, it’s so much easier to iterate on you probably want to exhaust caching-based strategies first.

[Marc Benioff @Benioff]

Build your own Gen AI Agents with Salesforce’s new Agentforce platform—anyone can create, test, and scale custom Gen AI Agents using our unique Einstein1 low-code platform—deeply integrated with our Einstein1 platform. The future of AI Agent and App Dev is in your hands!

With Agentforce, your custom AI agents:
- Achieve greater accuracy by securely grounding in all your enterprise data with Data Cloud
- Take action seamlessly by leveraging Flow automations
- Connect effortlessly to your enterprise APIs through MuleSoft
- Test with confidence in Sandboxes before deployment
- Use in your apps, web, social, and Slack.

Humans with AI drive customer success, together. Now everyone is an Einstein. More to come at Dreamforce 2024! 💙 https://t.co/wr0TmFkugj

[Jaya Gupta @JayaGup10]

We at @FoundationCap believe there is $4.6T of work to be automated. AI companies are leading a transition from Software-as-a-Service to Service-as-Software, turning the table on the very essence of SaaS.

We look at the areas to be automated in two buckets:

1.) Salaries of jobs globally ($2.3 trillion in sales & marketing, software engineering, security, and HR)

2.) The amount spent on outsourced services and salaries—both IT services and business process services ($2.3 trillion, per Gartner)

In the software business, a company may sell access to its platform or tool, but customers are still responsible for using that tool to achieve the desired outcome.

In the services business, responsibility for achieving the desired outcome sits with the company selling the service.

[Marc Benioff @Benioff]

Next-gen AI apps are just a click away for every Salesforce Trailblazer.
Craft AI that transforms enterprises with Salesforce! 🚀 Mix Copilot Builder, Model Builder, & Prompt Builder to revolutionize workspaces, supercharge employees, & empower customers.#AIRevolution #SalesforceMagic ✨🛠️

[Anthropic @AnthropicAI]

Today, we're announcing Claude 3, our next generation of AI models.

The three state-of-the-art models—Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku—set new industry benchmarks across reasoning, math, coding, multilingual understanding, and vision.

[Bill Gurley @bgurley]

This is what happens when you “invest” with “credits” that allow you to goose your own revenues. As long as that is allowed, expect it to continue and expect way more. And expect a massive mess in the end.

[Pietro Schirano @skirano]

I am FREAKING OUT. DesignerGPT is listed as number 1 in trending on OpenAI's official announcement for the store! Excuse me as I pass out!

[Omar Sanseviero @osanseviero]

E5 mistral-7b: New technique and SOTA model for text embeddings by Microsoft

Paper: https://t.co/CBC39lUl7f
Model: https://t.co/wA3QaiRdvi
Leaderboard: https://t.co/rIZ1pNPv1t

- Only trained on synthetic data (for 93 langs)
- Decoder-only LLM 🤯(Mistral 7B fine-tune)
- Tops MTEB and BEIR benchmarks (by quite a bit), including closed-sourced models
- Up to 32k tokens of input (rather than conventional 512 token limit)

Why is this interesting?
Long-context, high-quality embedding OS models are amazing for the community. Using a decoder model as a base seems somewhat an odd choice (as inference will be slower and size will be too large) that is having some attention (see RankLlama https://t.co/GgURmhI8VE). The fact that they purely use synthetic data is also quite exciting. It will be interesting to see how this evolves!

[Marc Benioff @Benioff]

🚀 Introducing Salesforce's game-changer: Data Cloud Vector Database + Einstein Copilot Semantic Search! ✨ Harness both structured & unstructured data for smarter AI responses. No more costly training! Real-time data-driven insights with RAG, complete with citations to curb AI hallucinations. Trust in enterprise AI just got a major upgrade, thanks to Salesforce! ❤️