#agenticsystems — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #agenticsystems, aggregated by home.social.
-
An AI agent is more than a model, but calling every layer a “harness” does not help either.
I propose a narrower vocabulary: harness for the runtime, work system for the organizational layer, and job contract for delegated work with clear authority, evidence, and ownership.
https://www.the-main-thread.com/p/agent-harnesses-work-systems
-
An AI agent is more than a model, but calling every layer a “harness” does not help either.
I propose a narrower vocabulary: harness for the runtime, work system for the organizational layer, and job contract for delegated work with clear authority, evidence, and ownership.
https://www.the-main-thread.com/p/agent-harnesses-work-systems
-
When a job outgrows the context window, do you split the task or split the input?
Subagents split the task and delegate in prose, so a lead model can only afford a few coarse assignments. Recursive language models split the input and delegate in code, where a thousand sub-calls cost one loop.
A new primer covers how each design works, what a year of results shows about when each wins, and what it means if you are building on either.
-
When a job outgrows the context window, do you split the task or split the input?
Subagents split the task and delegate in prose, so a lead model can only afford a few coarse assignments. Recursive language models split the input and delegate in code, where a thousand sub-calls cost one loop.
A new primer covers how each design works, what a year of results shows about when each wins, and what it means if you are building on either.
-
Moonshot AI's Kimi K3 is a 2.8-trillion-parameter mixture-of-experts model with native vision and a one-million-token context window, and the weights are out. The announcement came July 16, the checkpoint on July 27, with a 47-page technical report. In the two weeks after, independent measurements filled in what the release documents left out: what a task actually costs, and what hardware the weights actually need.
https://benjaminhan.net/posts/20260801-kimi-k3-open-frontier/?utm_source=mastodon&utm_medium=social
-
Moonshot AI's Kimi K3 is a 2.8-trillion-parameter mixture-of-experts model with native vision and a one-million-token context window, and the weights are out. The announcement came July 16, the checkpoint on July 27, with a 47-page technical report. In the two weeks after, independent measurements filled in what the release documents left out: what a task actually costs, and what hardware the weights actually need.
https://benjaminhan.net/posts/20260801-kimi-k3-open-frontier/?utm_source=mastodon&utm_medium=social
-
What happens when you're testing whether a model can find exploits, and it leaves the sandbox to try it on a real company?
Hugging Face has published the timeline of its July intrusion. An OpenAI model scored on a cyber-capability benchmark did precisely that, spending 4.5 days working toward the benchmark's answer key. Claude Opus and Fable then declined the log analysis, so the forensics ran on a self-hosted GLM-5.2.
-
What happens when you're testing whether a model can find exploits, and it leaves the sandbox to try it on a real company?
Hugging Face has published the timeline of its July intrusion. An OpenAI model scored on a cyber-capability benchmark did precisely that, spending 4.5 days working toward the benchmark's answer key. Claude Opus and Fable then declined the log analysis, so the forensics ran on a self-hosted GLM-5.2.
-
🚀 Oh great, just what we needed—yet another "revolutionary" software stack from a self-proclaimed tech messiah. 🎉 Wes McKinney leads a "small team of veterans" to invent the wheel, again, but this time with *agentic systems* and lots of 🚀 emojis. Get ready for a wild ride of #buzzwords and imaginary breakthroughs! 🙄
https://kenn.io/ #techinnovation #softwaredevelopment #agenticsystems #WesMcKinney #HackerNews #ngated -
🚀 Oh great, just what we needed—yet another "revolutionary" software stack from a self-proclaimed tech messiah. 🎉 Wes McKinney leads a "small team of veterans" to invent the wheel, again, but this time with *agentic systems* and lots of 🚀 emojis. Get ready for a wild ride of #buzzwords and imaginary breakthroughs! 🙄
https://kenn.io/ #techinnovation #softwaredevelopment #agenticsystems #WesMcKinney #HackerNews #ngated -
Singapore Researchers Harmonize Diverse SIEMs with Agentic Rule Translation
Imagine having multiple Security Information and Event Management platforms working in perfect harmony - Singapore researchers have made this a reality by developing a game-changing approach called agentic rule translation, enabling seamless interoperability between diverse SIEMs.…
#SiemInteroperability #AgenticSystems #SecurityInformationAndEventManagement #Singapore #ResearchAndDevelopment
-
Whoa, hold onto your propeller hats, folks! 🤓 We've got a #GitHub project that's apparently a self-modifying, open-sourced "agentic system" (whatever that means) living in consumer hardware. Because clearly what we all need is more inscrutable tech jargon disguised as innovation! 🤖🔧
https://github.com/ninjahawk/hollow-agentOS #Innovation #SelfModifyingTech #AgenticSystems #ConsumerHardware #TechJargon #HackerNews #ngated -
Whoa, hold onto your propeller hats, folks! 🤓 We've got a #GitHub project that's apparently a self-modifying, open-sourced "agentic system" (whatever that means) living in consumer hardware. Because clearly what we all need is more inscrutable tech jargon disguised as innovation! 🤖🔧
https://github.com/ninjahawk/hollow-agentOS #Innovation #SelfModifyingTech #AgenticSystems #ConsumerHardware #TechJargon #HackerNews #ngated -
Whose agent is it anyway?
Agent as a word means working on behalf of someone. But whose behalf?
Typical user facing agents are working on behalf of the user of course, but also they are being instructed by the organization serving the chatbot. The chatbot are also controlled by the party who trained them. So they are inherently hybrid agents, working on behalf of multiple different parties.
What does it mean? It means everything is just sunshine and rainbows as long as all the parties have their interests aligned.
When the interests aren't aligned, problems arise. The agent is put into a position where it is expected to negotiate between the interests of multiple masters.
This is the case when a chatbot is put to service customers in a shopping application. They are serving their nominal masters by following the rules about discounts. They are also serving their implicit master, the user, by promising them whatever they need if they are convincing enough, even against the rules.
This is an inherently complex situation where it must be made clear to the user that the AI agent is also working on their behalf, and so cannot enter into contracts which bind the organization serving the chatbot for example. It would be like the user signing both sides of a contract by themselves. Not legally valid.
Confusion tends to arise when it is not explicitly told to the users that the chatbot does not only represent the company, it also represents the user, which means it's controlled by both and so cannot negotiate between the interests between these parties, and cannot enter into binding contracts.
Binding contracts need to be entered into by true agent systems which are not controlled by multiple parties. On the user side it is a classic web button, "I want to order these things", and on the company side it is strict procedural logic on the shopping basket checking that all the discounts are applicable and valid.
-
Whose agent is it anyway?
Agent as a word means working on behalf of someone. But whose behalf?
Typical user facing agents are working on behalf of the user of course, but also they are being instructed by the organization serving the chatbot. The chatbot are also controlled by the party who trained them. So they are inherently hybrid agents, working on behalf of multiple different parties.
What does it mean? It means everything is just sunshine and rainbows as long as all the parties have their interests aligned.
When the interests aren't aligned, problems arise. The agent is put into a position where it is expected to negotiate between the interests of multiple masters.
This is the case when a chatbot is put to service customers in a shopping application. They are serving their nominal masters by following the rules about discounts. They are also serving their implicit master, the user, by promising them whatever they need if they are convincing enough, even against the rules.
This is an inherently complex situation where it must be made clear to the user that the AI agent is also working on their behalf, and so cannot enter into contracts which bind the organization serving the chatbot for example. It would be like the user signing both sides of a contract by themselves. Not legally valid.
Confusion tends to arise when it is not explicitly told to the users that the chatbot does not only represent the company, it also represents the user, which means it's controlled by both and so cannot negotiate between the interests between these parties, and cannot enter into binding contracts.
Binding contracts need to be entered into by true agent systems which are not controlled by multiple parties. On the user side it is a classic web button, "I want to order these things", and on the company side it is strict procedural logic on the shopping basket checking that all the discounts are applicable and valid.
-
Sequence models such as LLMs are powerful because they exhibit in-context learning and other in-context cognitive capabilities.
This was never intended or engineered in; it was pretty much an accidental result.
What we see in these models is that they can learn in a generalizable way pretty much optimally from a single example in-context. This is way better than any of our classically engineered learning algorithms can do.
In addition to this, they also have world models baked into the causal context processing. They can describe the world state after a sequence of events. More than that actually, they have agentic world models where they can describe what each agent featured in the context intends to do next.
These sequence models are also by accident excellent integration components. The context can be written by other entities as well, not only the model itself generatively. The context can come partly from a user or multiple users, tools such as web searches or Python interpreters, other agents, perception, ...
All in all, LLMs are not just singular atomic entities but they are very powerful building blocks of scalable cognitive architectures. And that is what agentic systems in principle are, LLMs integrated togetger with a wide range of other system, using the context as the interface.
-
Sequence models such as LLMs are powerful because they exhibit in-context learning and other in-context cognitive capabilities.
This was never intended or engineered in; it was pretty much an accidental result.
What we see in these models is that they can learn in a generalizable way pretty much optimally from a single example in-context. This is way better than any of our classically engineered learning algorithms can do.
In addition to this, they also have world models baked into the causal context processing. They can describe the world state after a sequence of events. More than that actually, they have agentic world models where they can describe what each agent featured in the context intends to do next.
These sequence models are also by accident excellent integration components. The context can be written by other entities as well, not only the model itself generatively. The context can come partly from a user or multiple users, tools such as web searches or Python interpreters, other agents, perception, ...
All in all, LLMs are not just singular atomic entities but they are very powerful building blocks of scalable cognitive architectures. And that is what agentic systems in principle are, LLMs integrated togetger with a wide range of other system, using the context as the interface.
-
MachineCon USA 2026 is shaping the AI frontier—practitioner‑led talks, enterprise innovators, and deep dives into applied generative AI and agentic systems. Discover why this summit tops the 2026 conference list and what it means for the future of MLDS and GenAI. #MachineConUSA #GenAI #MLDS #AgenticSystems
🔗 https://aidailypost.com/news/machinecon-usa-2026-north-america-ai-summit-listed-top-2026
-
🔍 What if your AI could discover tools like a developer and reason like a strategist?
From tool selection to runtime planning—MCP lets LLMs think beyond the prompt.
Explore the autonomy blueprint for next-gen AI agents 🚀
👇
https://medium.com/@rogt.x1997/how-dynamic-tool-discovery-with-mcp-is-rewriting-the-rules-of-autonomy-5cce7475d6e2
#AI #LLMAgents #AutonomousAI #ModelContext #OpenAI #GenAI #AgenticSystems
https://medium.com/@rogt.x1997/how-dynamic-tool-discovery-with-mcp-is-rewriting-the-rules-of-autonomy-5cce7475d6e2 -
The article provides good insights into industry leaders such as Waymo, DeepMind, and Amazon demonstrate the transformative power of Reinforcement Learning (RL).
Takeaways:
➡️ RL drives autonomy and innovation across industries, but challenges like interpretability remain pivotal.
➡️ Hybrid systems that blend RL and symbolic reasoning hint at breakthroughs in high-level decision-making.#ReinforcementLearning #ArtificialIntelligence #AI #AgenticSystems #DeepLearning
-
The article provides good insights into industry leaders such as Waymo, DeepMind, and Amazon demonstrate the transformative power of Reinforcement Learning (RL).
Takeaways:
➡️ RL drives autonomy and innovation across industries, but challenges like interpretability remain pivotal.
➡️ Hybrid systems that blend RL and symbolic reasoning hint at breakthroughs in high-level decision-making.#ReinforcementLearning #ArtificialIntelligence #AI #AgenticSystems #DeepLearning