@DSPyOSS

An open-source declarative framework for building modular AI software. Programming—not prompting—LLMs via higher-level abstractions & optimizers.

dspy.ai
Joined April 2025
  • Tweets3.4K
  • Following64
  • Followers14.5K
92Media
applications for slingshots more than doubled from batch 3 to batch 4. funding open science is critical to democracy. never in the past have scientists had to turn down so much money to do research they can talk about openly. top PhD students are receiving offers of millions of dollars per year from closed labs. i'm proud that at Laude we support open research. great work on slingshots batch four @bradenhancock and the @laudeinstitute team
Laude Institute
@LaudeInstituteOct 20
Introducing Slingshots // FOUR: 14 projects across eight top institutions. Self-improving agent systems and harnesses, maximally-open models and training data, applications for research and decision-making – plus kernels and more. Say hello to the batch 🧵
2
8
1
48
4,490
I spent a little time last weekend swapping the sub-calls inside Recursive Language Models for Jev, that cute lil' System One model that returns calibrated probabilities. At 1M tokens it's half the cost and 6× faster than Haiku 5.5 on OOLONG. Also some fun findings on abstention, routing to bigger models, and where it didn't help at all 🧵
4
5
30
2,796
@DSPyOSS GEPA beat autoresearch on optimizing a skill to make Claude write like me. @LakshyAAAgrawal
2
14
2,199
The book is in pre-release here, and should be on Amazon soon! oreilly.com/library/view/con…

Context Engineering with DSPy

AI agents need the right context at the right time to do a good job. Too much input increases cost and harms accuracy, while too little causes instability and hallucinations.... - Selection from...

oreilly.com
3
6
2
42
8,386
I got a fish this time 🐟
8
13
3
117
15,467
Small @DSPyOSS bump with a new @gepa_ai proposer (add instructions, set length limits, compact traces), DSPy in RLM subagents, and plenty of bug fixes and 1st time contributors! github.com/stanfordnlp/dspy/…

Release 3.4.1 · stanfordnlp/dspy

DSPy 3.4.1 DSPy 3.4.1 makes GEPA's structured InstructionProposer the default instruction proposer and lets dspy.RLM build DSPy sub-agents from inside its REPL. It also adds Python 3.15 support...

github.com
3
13
62
4,135
Replying to @AlexSlobodnik
that's why i love the direction of @DSPyOSS, given the extra layer of structuring json for instructions and criteria especially (which is powerful but volatile technique), set it and forget it with GEPA or dspy.Flex is a really cool approach
1
2
10
736
I’ll be in San Francisco this week and also attending the Simons workshop! If you're around and want to catch up, please reach out. I'd love to chat about Pedagogical RL, diversity, alignment and anything beyond standard RL.
8
5
1
68
13,559
DSPy retweeted
Alec
@AlecTPhDOct 3
Replying to @StijnSmits @a1zhang @willcb @lateinteraction
Yep! Claude Science has a python repl with a host bridge that gives it access to harness primitives so it can do lots of things programmatically. MCP tool calling, skill creation, agent profile creation, LLM calling, delegation, and more :) The python repl is also perfect for doing computational science since bulky datasets stay in memory as agents iterate on their analysis!
9
19
1
221
21,562
DSPy retweeted
Stijn
@StijnSmitsOct 3
AI Engineer at Anthropic just exposed Claude Science is an RLM 🚨 @a1zhang @willcb @lateinteraction
18
16
2
377
78,308
dspy.Flex primitive will literally write your pipeline code for you, helping with de-composition. dspy.ai/current/api/modules/…
2
5
42
3,049
Replying to @typesafeai @DSPyOSS
@dotpem has an excellent breakdown of the highlights xitter.cf/dotpem/status/21054179…
Nathan LeClaire
@dotpemSep 30
the @DSPyOSS town hall about dspy+jev was really good. here are some of my favorite highlights
1
9
7,354
This office hours recording from @DSPyOSS is a gold mine of excellent Jev use-cases and explainers!
Drew Breunig
@dbreunigSep 30
Today at 2pm PT, @isaacbmiller1 and I will be walking through @DSPyOSS's Jev/System One implementation & the ReAnchor optimizer. Will also sharing some design patterns for for selective compaction, tool approvals, and subagent delegation using Jev. streamyard.com/watch/sbMTpJA…

DSPy Office Hours (with Jev!)

Walking through DSPy's Jev integration, the ReAnchor optimizer, and compound AI design patterns with Jev.

streamyard.com
4
11
158
25,517
Replying to @Keleesssss
DSPy is one such language specifically tailored to functions with natural language based operations. DSPy is built around a “signature” which has inputs and outputs and a natural language specification of the task. Then we provide structure to define what “good” looks like, and optimize the implementation underneath this function signature. Here, rather than strictly unit tests as is usually done in swe, we use a dataset/eval. At first, we would only optimize the few shot examples of any of these functions, but as language models have gotten better we have moved to optimizing the prompts and now the code which completes your signature. You might like a blog from @michaelnisaac on the Module which lets you learn code, DSPy.flex cmpnd.ai/blog/let-the-model-…

Introducing Flex: Let the Model Write the Code — Cmpnd

Introducing Flex, a DSPy module that exposes its own source code to the optimizer. On a location conflation task, GEPA rewrote the program to be more accurate, 28% cheaper, and 40% faster than the...

cmpnd.ai
1
6
33
1,393
Last year I made a presentation on @DSPyOSS on @PyDataGlobal and I was thinking making it into a book, as there were basically no books on it. Bad news: I had no time to write the book. Great news: Manning Publications contacted me to review this book, which was really well done. Focused on production. Now I don't have to do it. Thanks for the invitation and congrats on the launch, Serj and Brett!
1
5
402
the @DSPyOSS town hall about dspy+jev was really good. here are some of my favorite highlights
2
11
1
70
18,620
DSPy Office Hours (with Jev!) xitter.cf/i/broadcasts/1aKbdEvyP…
0:00

Drew Breunig

DSPy Office Hours (with Jev!)

1
4
35
1,741
Come discuss how to use Jev in DSPy!
Drew Breunig
@dbreunigSep 28
This Wednesday at 2PM Pacific, @isaacbmiller1 and I will be walking through DSPy's Jev integration, the ReAnchor optimizer, and discussing a few design patterns. (Including some detailed by @CompleteSkeptic in his recent post on agents) Join us! streamyard.com/watch/sbMTpJA…

DSPy Office Hours (with Jev!)

Walking through DSPy's Jev integration, the ReAnchor optimizer, and compound AI design patterns with Jev.

streamyard.com
1
4
21
3,088
Fantastic talk from @isaacbmiller1 tonight about using DSpy with Jev to build AI programs. Really interesting use of DSpy to learn the optimal confidence threshold.
2
6
80
4,208
DSPy retweeted
Diane
@dianetc_Sep 28
I wrote a short blog on my experience testing Jev on reasoning-intensive regression problems. The results are within expectations (Jev is focused on classification after all), but the model has great attributes and is wicked fast! Read about it here: dianetc.com/musings/jev/
2
13
1
105
21,502
Load more