Tweets

Philipp Schmid

5d ago

Proposal: Agent Hooks 🏗️

1️⃣ mount: Initial setup/config, tool definition 
2️⃣ before: Dynamic context & SI changes
3️⃣ after: Human in the Loop / Safety

I found similar in @LangChainAI with Middleware and Vercel @aisdk with call options.

I’m going to try this out. Instead writing tools and have 1 SI, I’ll write Hooks that can dynamically modify System Instructions, Compress context, or inject specific tools based on the conversation state.

Philipp Schmid

5d ago

Feels like tool(sets) and MCP server need a better way to append to system instructions for combinations

Example: 
'if you use the GitHub to create a PR, always run the Linear tool to create a linked task afterward’. But this makes only sense if you have both MCP servers/tools in the scope.

113

26.5K

147

109

17.4K

Google AI Developers

2d ago

Gemini 3, the state of AI, how we’re approaching building AGI, and much more.

Check out this conversation between @koraykv (@Google’s Chief AI Architect and @GoogleDeepMind CTO) and @OfficialLoganK ↓
x.com/GoogleAI/statu…

Google AI

1w ago

A LOT happened last week. ICYMI, @koraykv (CTO of @GoogleDeepMind and Chief AI Architect of @Google) and @LoganKilpatrick discuss Gemini 3, the state of AI, and where we are on the path to AGI.

Chapters:
0:00 - Intro
2:00 - Gemini 3 launch reception
4:16 - Continuous progress and innovation
6:47 - Key areas for Gemini improvement
11:45 - Product scaffolding for model improvement
13:56 - Chief AI architect role
17:04 - Engineering mindset and collaboration
18:37 - Future growth areas for Gemini
20:33 - From research to engineering mindset
23:22 - The rise of generative media
27:22 - Nano Banana Pro capabilities
29:31 - Towards unified model checkpoints
36:26 - Organizing for AI success
38:26 - Balancing exploration and scaling
41:40 - DeepMind's collaborative culture
45:21 - Innovating at Google
48:37 - Closing

48:44

444

151.3K

251

19.3K

Harrison Chase

1d ago

🚀 LangSmith Agent Builder in Public Beta

Agent Builder (now available for every LangSmith user) opens the gates to everyone to build agents without code

Video on why no-code agents can open up agent building and how I built a recruiting agent with it:youtu.be/6-87bwBCyos4

00:56

13.4K

Y Combinator

2d ago

.@inspectmind's AI agent reviews construction drawings, specs, and building codes to catch issues before they turn into rework, delays, or lawsuits.

It flags dimension errors, materials mismatches, code violations, and coordination gaps across architectural, structural, civil, and MEP drawings.

On one $100M job, it caught a materials mismatch that saved the general contractor $2M.

ycombinator.com/launches/OtS-i…

01:31

16.3K

OpenRouter

2d ago

📢 @arcee_ai's new Trinity Mini is live on OpenRouter!

Trinity Mini is a 26B param MoE model end-to-end trained in the United States that delivers efficient long-context inference, strong multi-turn reasoning, and reliable tool calling at scale.

Lucas Atkins

2d ago

Today, we are introducing Trinity, the start of an open-weight MoE family that businesses and developers can own.

Trinity-Mini (26B-A3B)
Trinity-Nano-Preview (6B-A1B)

Available Today on Huggingface.

00:45

154

1,313

597.2K

109

14.3K

Thomas Wolf

1d ago

Excited to see the open-source lab @kyutai_labs spinning out @GradiumAI – increasingly growing potential in robotics and real-time applications

Gradium

1d ago

Gradium is out of stealth to solve voice. We raised $70M and after only 3 months we’re releasing our transcription and synthesis products to power the next generation of voice AI.

02:23

135

968

331.5K

11.2K

Jerry Liu

1d ago

We've been cooking 🧑‍🍳

📊 Spreadsheet parsing (LlamaSheets) 
📑 Multi-step document workflows (LlamaAgents)
❓ 3 new releases coming out this week and next week 

If you have core use cases around document processing, we'd love to talk to you. Come check out our office hours hosted by @tuanacelik!

Tuana

1d ago

The team at @llama_index have been cooking! 🧑‍🍳 🍳
Over the last few weeks, we released:

LlamaAgents: This is agent workflows that come with complete, deployable templates (more coming on this this week!)
LlamaSheets: Another addition to LlamaCloud that parses, extracts information and deep context (also hidden in metadata) from tables and spreadsheets, as well as identifying sub-groups from complex sheets

That's a lot, so this week, we want to take the time to hear from you! On Thursday, me, @LoganMarkewich and @itsclelia will be in our Discords voice channel for an hour dedicated to chatting about these two new tools. Drop by for our office hours, we'd love to hear from you. And of course, bring along any questions too!

5,773

4,733

Alex Albert

1d ago

anthropic.com/news/anthropic…

5,566

Google AI Developers

2h ago

For complex agents, enforcing behaviors like persistence, risk assessment, and proactive planning drives better reliability in multi-step workflows.

Resources 🔗 
SI Template:ai.google.dev/gemini-api/doc…w
Docs:ai.google.dev/gemini-api/doc…U

1,714

Aadit Sheth

6d ago

AI closes the gap on doing.

Taste closes the gap on deciding.

And deciding is where all the value is created.

AI will take over the typing.

It won’t take over the knowing.

3,331

Sources

Proposal: Agent Hooks 🏗️

1️⃣ mount: Initial setup/config, tool definition
2️⃣ before: Dynamic context & SI changes
3️⃣ after: Human in the Loop / Safety

I found similar in @LangChainAI with Middleware and Vercel @aisdk with call options.

I’m going to try this out. Instead writing tools and have 1 SI, I’ll write Hooks that can dynamically modify System Instructions, Compress context, or inject specific tools based on the conversation state.

Feels like tool(sets) and MCP server need a better way to append to system instructions for combinations

Example:
'if you use the GitHub to create a PR, always run the Linear tool to create a linked task afterward’. But this makes only sense if you have both MCP servers/tools in the scope.

Gemini 3, the state of AI, how we’re approaching building AGI, and much more.

Check out this conversation between @koraykv (@Google’s Chief AI Architect and @GoogleDeepMind CTO) and @OfficialLoganK ↓
x.com/GoogleAI/statu…

A LOT happened last week. ICYMI, @koraykv (CTO of @GoogleDeepMind and Chief AI Architect of @Google) and @LoganKilpatrick discuss Gemini 3, the state of AI, and where we are on the path to AGI.

Chapters:
0:00 - Intro
2:00 - Gemini 3 launch reception
4:16 - Continuous progress and innovation
6:47 - Key areas for Gemini improvement
11:45 - Product scaffolding for model improvement
13:56 - Chief AI architect role
17:04 - Engineering mindset and collaboration
18:37 - Future growth areas for Gemini
20:33 - From research to engineering mindset
23:22 - The rise of generative media
27:22 - Nano Banana Pro capabilities
29:31 - Towards unified model checkpoints
36:26 - Organizing for AI success
38:26 - Balancing exploration and scaling
41:40 - DeepMind's collaborative culture
45:21 - Innovating at Google
48:37 - Closing

🚀 LangSmith Agent Builder in Public Beta

Agent Builder (now available for every LangSmith user) opens the gates to everyone to build agents without code

Video on why no-code agents can open up agent building and how I built a recruiting agent with it:youtu.be/6-87bwBCyos4

.@inspectmind's AI agent reviews construction drawings, specs, and building codes to catch issues before they turn into rework, delays, or lawsuits.

It flags dimension errors, materials mismatches, code violations, and coordination gaps across architectural, structural, civil, and MEP drawings.

On one $100M job, it caught a materials mismatch that saved the general contractor $2M.

ycombinator.com/launches/OtS-i…

📢 @arcee_ai's new Trinity Mini is live on OpenRouter!

Trinity Mini is a 26B param MoE model end-to-end trained in the United States that delivers efficient long-context inference, strong multi-turn reasoning, and reliable tool calling at scale.

We've been cooking 🧑‍🍳

📊 Spreadsheet parsing (LlamaSheets)
📑 Multi-step document workflows (LlamaAgents)
❓ 3 new releases coming out this week and next week

If you have core use cases around document processing, we'd love to talk to you. Come check out our office hours hosted by @tuanacelik!

The team at @llama_index have been cooking! 🧑‍🍳 🍳
Over the last few weeks, we released:

LlamaAgents: This is agent workflows that come with complete, deployable templates (more coming on this this week!)
LlamaSheets: Another addition to LlamaCloud that parses, extracts information and deep context (also hidden in metadata) from tables and spreadsheets, as well as identifying sub-groups from complex sheets

That's a lot, so this week, we want to take the time to hear from you! On Thursday, me, @LoganMarkewich and @itsclelia will be in our Discords voice channel for an hour dedicated to chatting about these two new tools. Drop by for our office hours, we'd love to hear from you. And of course, bring along any questions too!

For complex agents, enforcing behaviors like persistence, risk assessment, and proactive planning drives better reliability in multi-step workflows.

Resources 🔗
SI Template:ai.google.dev/gemini-api/doc…w
Docs:ai.google.dev/gemini-api/doc…U