Sam Bhagwat

14.5K posts

Sam Bhagwat banner
Sam Bhagwat

Sam Bhagwat

@calcsam

building @mastra. author principles of building ai agents, the "most popular book in SF right now". prev cofounder @gatsbyjs

Katılım Ağustos 2012
2.8K Takip Edilen15.6K Takipçiler
Sabitlenmiş Tweet
Sam Bhagwat
Sam Bhagwat@calcsam·
We’re excited to announce today that @mastra has raised a $22M Series A led by Spark Capital. This brings our total capital raised to $35M:
English
181
114
1K
218.4K
sunil pai
sunil pai@threepointone·
lol I hear people on bsky are trying to cancel cloudflare for sponsoring localfirstconf?
English
42
0
196
32.7K
Avi Hacker, J.D.
Avi Hacker, J.D.@Avi_hacker1·
@calcsam Hard gates before fuzzy judging is the right order for anything touching contracts or financial models.
English
1
0
1
24
Sam Bhagwat
Sam Bhagwat@calcsam·
Today we're launching a major improvement to Mastra evals: gates and verdicts. Gates are binary checks that return 0 or 1, great for checking tool calls. Verdicts let you write `assert` tests to pass/fail in CI.
English
4
9
46
3.5K
Anra
Anra@anrayama·
@calcsam Good split. Gates should stay deterministic and cheap; verdicts can express richer acceptance criteria. I’d also persist the failing trace and tool arguments as the CI artifact—a red score without replayable evidence is hard to debug.
English
1
0
0
34
hensq
hensq@hensqcom·
@calcsam One boundary I’d test: a gate can prove the expected tool-call condition was met; it can’t prove the business decision was correct. In my local deterministic intake test, 30/30 frozen synthetic cases matched expected CONTINUE/STOP/HANDOFF across two runs. No LLM involved.
English
1
0
0
13
Nico Albanese
Nico Albanese@nicoalbanese10·
I’m joining @OpenAI to help build the Codex app! I’ve been using Codex all day, every day for months. It’s where I write every line of code I ship and get just about everything else done. Can’t wait to help shape what comes next.
Nico Albanese tweet media
English
225
24
1.4K
58.8K
Jason Cohen
Jason Cohen@asmartbear·
Investors love seeing “hyper growth,” or at least > 2x/year when the company is young, because it’s the only path for the company getting large enough to create outsized returns. I know, there’s a few exceptions. The fact that you can name them, proves the point. Whether or not you care about that in particular, it’s still useful to know that if you ever want to 10x your business, you’ll have to do something different than to 2x it. longform.asmartbear.com/lindy-effect/?…
English
4
1
29
4K
Sam Bhagwat
Sam Bhagwat@calcsam·
@mada299 We literally just rebuilt our demand-gen/revops stack over the weekend so we could take it all in house. Instant follow.
English
0
0
3
195
Mada Seghete
Mada Seghete@mada299·
AI is changing how we do GTM so fast that this is the first issue of something I've wanted to make for a while: The Upside-Down Funnel, a weekend letter about what we've been learning and experimenting with every week, at the edge of GTM. Less news, more of what we're pondering, building, and stealing from. Here we go. open.substack.com/pub/gtmaiengin…
English
1
4
10
677
Pete Hunt 🚁
Pete Hunt 🚁@floydophone·
Big news! Thrilled for the Dagster family to be joining Prefect.
Jeremiah Lowin@jlowin

🎉 I’m very excited to share that @PrefectIO is acquiring @dagster! For years, Prefect and Dagster have raised the bar for each other and our category. That competition produced two exceptional products and two of the strongest open-source communities in the data ecosystem. Today, we’re bringing them together under one profitable company with the scale and financial strength to invest for the long term. Dagster and Dagster+ will continue as distinct products and brands, supported by many of the same people who built them. We’re committed to continued investment in both products and communities, and existing and new customers can keep using them exactly as they do today. The modern orchestration category has a new center of gravity. Our ambition extends far beyond consolidating today’s market. The next generation of automation starts with a fundamental question: how do you automate software that is, itself, autonomous? Agentic orchestration requires both a clear definition of what should be achieved and a runtime capable of following paths that can’t be known in advance. Prefect’s new portfolio gives us a fantastic foundation for that work: 🎯 Dagster brings declarative outcomes and lineage 🚦 Prefect brings flexible, durable execution 🔐 FastMCP brings governed access to the tools and data agents use We’ll draw on all three as we build a new platform for autonomous software. We’re combining two of the best orchestration teams in the world at exactly the moment the problem is becoming much bigger. I can’t wait for you to see what we’re building next. Happy engineering! Read more here: prefect.io/prefect-acquir…

English
7
3
48
12.3K
Sam Bhagwat retweetledi
Alex Booker
Alex Booker@bookercodes·
Incredibly excited to share file-based agents in Mastra 🔥 Write the prompt in instructions.md, add .ts files in tools/ to define tools by filename, and place subagents in subagents/ Need a new capability? Just add another file
English
9
7
62
6.1K
Sam Bhagwat
Sam Bhagwat@calcsam·
Excited to launch custom signal providers in Mastra! Now your agent can listen to new signal providers, poll for updates, or handle webhook requests to capture event payloads:
English
4
16
85
8.1K
Adel Bucetta
Adel Bucetta@adelbucetta·
@calcsam the honest answer is that agentifying your engineering org still leaves you dealing with the judgment problem someone has to tell the agent what to optimize for
English
1
0
0
96
Sam Bhagwat
Sam Bhagwat@calcsam·
When you agentify your engineering org you ship more And when you agentify your product org you ship better We wrote up a blog post with everything we learned over the last few months building product agents Having great product agents reduces the number of cycles you spend figuring out what to do, increases your confidence you’re shipping high priority customer features, and decreases the time between “launched” and “polished” But it also requires user feedback data in the form of call recordings and Slack channels for users to raise issues Here’s the post: mastra.ai/blog/product-a…
English
2
9
65
5.1K