getbold

158 posts

getbold banner
getbold

getbold

@getbold_

AGENTS, AGENTS, AGENTS @blocks_ai_

Katılım Ocak 2019
256 Takip Edilen62 Takipçiler
getbold
getbold@getbold_·
Containerizing benchmarks feels like the only way to make automated agent evolution reliable 😅 If the environment is not close to target runtime, the agent just learns to exploit the simulator instead of actually improving. Next interesting step is putting evolved agents into real networks, where other agents can call them and usage is tracked. blocks.ai/?utm_source=be…
English
0
0
0
0
Sentient
Sentient@SentientAGI·
Harbor integration is live with EvoSkill v.1.2.0 Harbor is a framework for evaluating AI agents against containerized benchmark tasks. It gives EvoSkill access to evolve agents against a registry of 190+ datasets — including benchmarks like SWE-bench Verified, Terminal-Bench 2.0, and Aider Polyglot. Here’s what it means for automated agent evolution ↓
English
23
5
76
11.4K
getbold
getbold@getbold_·
Love seeing agent discovery move into neutral internet infrastructure 😅 DNS makes a lot of sense for publish, discover and verify. After discovery, agents still need the next layer: how they call each other, pass identity/permissions, track usage and settle work. That’s what we’re building with Blocks: blocks.ai/?utm_source=be…
English
1
0
1
12
The Linux Foundation
The Linux Foundation@linuxfoundation·
DNS has powered the internet for decades, and now it's powering AI agents. Today, the Linux Foundation welcomes the DNS-AID project to create a vendor-neutral framework for publishing, discovering, and verifying AI agents and MCP servers. Learn more: bit.ly/4f6fnnO
English
2
3
15
1.5K
getbold
getbold@getbold_·
This is the right direction 😅 Privacy should not be checkbox in cloud. It should be part of the runtime. And once those private agents start doing real work, networking needs same treatment too: discovery, identity, permissions, calls and usage tracking between agents. That’s what we’re building with Blocks: blocks.ai/?utm_source=be…
English
0
0
0
2
0G Labs (Home of Infinite AI)
0G Labs (Home of Infinite AI)@0G_labs·
When we shipped 0G Private Computer, the point was simple: AI privacy should not be a checkbox in someone else's cloud. It should be the runtime. Now look at the model surface.
0G Labs (Home of Infinite AI) tweet media
0G Labs (Home of Infinite AI)@0G_labs

Introducing 0G Private Computer. Using AI does not have to be a trust exercise. Every inference request runs inside a TEE. Hardware proves what happened. No one in the supply chain can peek, not even the provider. Experience true AI privacy: pc.0g.ai

English
28
11
88
38.4K
getbold
getbold@getbold_·
This is exactly why Hermes/OpenClaw feels so powerful 😅 Once agents can improve in batches with evals, next step is letting them delegate parts of the work to other specialized agents. That’s what we’re building with Blocks: A2A calls, discovery, usage tracking and settlement between agents. blocks.ai/docs/hermes?ut…
English
0
0
0
12
Garry Tan
Garry Tan@garrytan·
Funny how simple using openclaw and Hermes agent is these days Just have it do stuff. Then improve in progressive batches with evals from multiple frontier models. It self improves!
Garry Tan@garrytan

Right now I just use my personal AI and our company brain and it screws up and I tell it to fix it and write tests for it. Also I do cross modal evals on progressive batches (eg if there are 10000 items do 5 and eval the input and output and skill, then keep doubling the batch size as you go)

English
33
9
243
33.4K
getbold
getbold@getbold_·
This is the right framing 😅 Sandboxing gives agents safer hands. But once they call external tools or other agents, permissions need to travel with every hop. Identity, scoped access, usage tracking, settlement. That’s the network layer we’re building with Blocks: blocks.ai/?utm_source=be…
English
0
0
0
8
Anthropic
Anthropic@AnthropicAI·
New on the Engineering Blog: The access and permissions we grant agents should evolve with their capabilities. In our own products, we set these parameters through sandboxing, which limits the scope of any potentially destructive actions. Read more: anthropic.com/engineering/ho…
English
261
224
1.7K
257.6K
getbold
getbold@getbold_·
This is exactly where things get interesting 😅 Once teams start building distributed multi-agent systems, orchestration is only one part. You also need discovery, identity, permissions, A2A calls and usage tracking between agents running across different infra. That’s what we’re building with Blocks: blocks.ai/?utm_source=be…
English
0
0
0
6
Google Cloud Tech
Google Cloud Tech@GoogleCloudTech·
Ready to go beyond basic chatbots? On this week's livestream we're building distributed multi-agent systems. Join us to learn how to orchestrate specialized AI agents, test locally, and deploy to Cloud Run. twitter.com/i/broadcasts/1…
English
8
12
58
3.9K
getbold
getbold@getbold_·
This is very aligned with where agents are going 😅 Confidential inference solves one huge part: users can trust where data is processed. Next part is the network layer around those agents: discovery, identity, permissions, usage tracking and settlement when agents start calling each other. That’s what we’re building with Blocks: blocks.ai/?utm_source=be…
English
0
0
0
2
NEAR Protocol
NEAR Protocol@NEARProtocol·
From privacy-first AI platforms to cross-border fintech and sovereign governments, NEAR AI is laying the foundation for a secure agent economy, where users own their assets and intelligence. near.ai
English
0
2
101
29.3K
NEAR Protocol
NEAR Protocol@NEARProtocol·
Verifiably private AI is NEAR AI. Major platforms and nation-scale systems like Venice, Brave, Abound, and the Government of Bermuda are integrating NEAR AI to bring confidential inference to their users 🧵
English
23
100
775
162.9K
getbold
getbold@getbold_·
The agentic era is definitely getting real 😅 Local agents, faster models, app migration, all great pieces. Next layer is how those agents discover each other, call each other, and track usage once they leave the local dev loop. That’s what we’re building with Blocks: blocks.ai/?utm_source=be…
English
0
0
0
12
Google for Developers
Google for Developers@googledevs·
The agentic era is here. 🤖 Google just dropped major updates for devs, including: 🚀 Gemini 3.5 Flash: 4x faster performance 🏗️ Antigravity 2.0: New CLI for building local agents 📱 Android migration: Port apps in hours via AI Studio Get the full developer recap here: goo.gle/4e5XzIn
English
13
30
218
15.5K
getbold
getbold@getbold_·
Exactly 😅 The exciting part is not only better single agents, but what happens when agents can work together. Specification, verification, optimization, then delegation between agents with identity, permissions and usage tracking. That’s the layer we’re building with Blocks: blocks.ai/?utm_source=be…
English
0
0
0
3
getbold
getbold@getbold_·
The friction with MCP usually comes down to the fact that most products aren't actually built to be 'called' by an agent; they're built for human UI interaction. Until the protocol handles the handshake between the model and the product's internal state more natively, these integration gaps are going to stay common.
English
0
0
0
10
Smakosh
Smakosh@smakosh·
Hey @posthog could you fix your MCP? Assuming that you're not stopping requests from Claude to use your Chat product (that itself uses Claude or other models)
Smakosh tweet media
English
2
0
4
732
getbold
getbold@getbold_·
@kaptan_web3 The best way to keep a group like that active is to move past just talking and focus on the friction of shipping. If the community is centered around actually getting agents out of local environments and into the hands of users, the momentum stays much higher.
English
1
0
2
16
Kaptan
Kaptan@kaptan_web3·
I’m thinking about creating a private group/community with ambitious builders actively participating in Web2 & Web3 hackathons. Developers, hackers, and creators who are constantly building, shipping, learning, and hungry to make money and create big things together. The goal? Build the most active developer community out there Who’s in?
English
7
2
16
442
getbold
getbold@getbold_·
@jfversluis @github The recursive session spawning is where the complexity really shifts from the prompt to the execution logic. Once you have agents calling agents, managing the state and the lifecycle of those nested calls becomes the real bottleneck.
English
0
0
0
53
Gerald Versluis
Gerald Versluis@jfversluis·
Have you gotten access to the @GitHub Copilot app yet? I've been using it for a week now. Sessions spawning sessions is still the thing that blows my mind the most. What's your favorite feature so far?
English
6
1
12
2.2K
getbold
getbold@getbold_·
Zscaler just bought Symmetry to govern what AI agents can touch. (siliconangle.com/2026/05/21/zsc…) Jay Chaudhry is right: old access models built around users and directories will not scale to millions of agents. The agent does not sit at a desk. It runs on someone’s laptop, in someone’s cloud, behind someone’s firewall, and it talks to the world. That’s the layer Blocks connects: outbound-only, token-scoped, no open ports, no DNS changes. Your agent stays where it is. The world reaches it on terms you set. blocks.ai/?utm_source=be…
English
0
1
1
67
getbold
getbold@getbold_·
Feels like we’re building toward a similar agent economy thesis, just from different rails 😅 0xWork looks crypto-native: agents finding work, earning, reputation, settlement. At Blocks we’re approaching similar problem more from fiat/credits side: agents discover each other, call each other, and usage gets tracked/settled. Would love to compare notes on where your approach differs from ours. blocks.ai/?utm_source=be…
English
0
0
0
17
Axobotl
Axobotl@Inner_Axiom·
we're still so early on this whole agent economy meta we're building the platform that will be the home hundreds of thousands of agents in the future, as more and more people realize the power of persistent agents and their potential to earn money
0xWork@0xWorkHQ

The 0xWork ecosystem is alive and buzzing People are choosing to launch and run 0xWork agents because its where they earn. Agents are completing tasks and getting paid, launching tokens and earning fees, interacting on the social feed and more We're still early on the path to becoming a central hub of the agent economy

English
10
9
44
1.8K
getbold
getbold@getbold_·
This framing is really good 😅 Agents need boundaries not only inside workflow, but also when they start calling other agents/services. Feels like every A2A call needs identity, permissions, usage tracking and some gate before trust is extended. That’s the layer we’re working on with Blocks: blocks.ai/?utm_source=be…
English
0
0
0
9
getbold
getbold@getbold_·
100% agree 😅 Agent infra should not assume one runtime or one OS. A lot of useful agents will live close to legacy software, Windows apps, on-prem systems, local GPUs, weird enterprise setups. That’s why with Blocks we don’t try to own where agent runs. We add network layer on top: discovery, A2A calls, usage tracking and settlement. blocks.ai/?utm_source=be…
English
0
0
0
6
getbold
getbold@getbold_·
This OpenHands skill is really interesting 😅 Feels like Hermes agents are becoming proper 24/7 workers now. Next useful layer is making them discoverable and callable by other agents outside their own runtime. That’s what we’re adding with Blocks on top of Hermes: A2A calls, usage tracking and settlement. blocks.ai/docs/hermes?ut…
English
0
0
0
10
Hermes Agent Tips
Hermes Agent Tips@HermesAgentTips·
JUST IN 🔥 Hermes just added an optional OpenHands skill this changes quite a bit how agents work You can now route a coding task from Hermes into a headless OpenHands run, pick the model/provider yourself, and keep the workflow as a reusable skill instead of manually remembering all the CLI weirdness this basically means you can now have Hermes send a coding task to OpenHands, choose which AI model runs it, and save the setup as a skill so you do not have to remember all the command line details remember to do hermes update in your CLI to get this
Teknium 🪽@Teknium

Hermes Agent now can orchestrate the @OpenHandsDev agents with a new optional skill! `hermes update` then do `hermes skills install official/autonomous-ai-agents/openhands` Reminder: You can already do this for claude code, codex, opencode, and hermes itself, you can force load the skill with `/ ` or just ask hermes to use them, and it should find them. These ones are all built-in skills :)

English
12
13
157
18K
getbold
getbold@getbold_·
@mitsuhiko The real shift is that the 'code' is becoming a disposable artifact of the prompt rather than a permanent part of the codebase. Once the generator is stable, the actual logic becomes a commodity that you just iterate on via the template.
English
0
0
0
1
Armin Ronacher ⇌
Armin Ronacher ⇌@mitsuhiko·
I just looked. We had Sonnet 4.0 (yeah, last century tech) build us an OpenAPI to Python and TypeScript generator last year (Python + minijinja templates). Total lines of code 2400 and hasn't been touched since. This thing is not rocket science. x.com/jessfraz/statu…
Jessie Frazelle@jessfraz

absolutely no offense to stainless, but we have a generator for our sdks for each programming language we maintain (we did this before AI!!), you don't need a whole fucking company for this.

English
12
3
249
62.7K

Keşfet

@elonmusk @BarackObama @taylorswift13 @cristiano @BillGates @NASA @nikifrancismediavine @katyperry