Superflux AI

[Superflux AI @SuperfluxSF]

So far practical coding with Astra has been underwhelming. Tends to invent scope and requirements. I am not giving up yet. Just complaining. So far the benchmarks have not showed up in my coding outcomes.

[Superflux AI @SuperfluxSF]

Clop aka Slop from Claude Opus 5

"The engine being on makes the MEDIUM change reachable. It doesn't make it reached"

"You asked for prose. What you got was poetry with tautologies"

"He’s right that it’s a real problem. I was wrong to wave it off. But the mechanism is neither his nor mine, and that’s what changes the fix." (in response to the independent review subagent")

[Julie Yoo @julesyoo]

https://t.co/GvSA9ykYP8

[Nous Research @NousResearch]

Hermes Agent now runs Buzz.

The self-hostable workspace from @blocks puts humans and agents in the same messaging channels and codebase.

Three ways to use Buzz with Hermes (and vice versa):
- Buzz Desktop auto-discovers your Hermes install runs it locally
- A relay bridge gives it a hosted identity in your channels
- Connect via the Hermes Gateway to use Buzz as a full external platform with channels, DMs, threads, reactions, and cron delivery

https://t.co/srljIbERN7

[Superflux AI @SuperfluxSF]

@ronxjansen few things. random screening - peek at the running agent commentary and steer, or kill the run (doesnt work AFK). In rare cases, just throwaway and restart Highest leverage - drift check skill on every commit hook and pause goal.

[Superflux AI @SuperfluxSF]

Codex agents can run for a very long time (days) when you design your factory right. Here are the key ingredients

1. Correct harness setup
The right model, AGENTS.md, worklog.md, ADRs, and plugins like Compound Engineering, Ambiguity Filter, Playback, Cognitive Render, Change Management, PR Babysitter, Drift Detector, LFDD, TDD, and EDD.

2. Human-agent alignment layer
Start strong. Remove ambiguity. Align on the specs in gory detail. Then align on the technical plan in gory detail.

Plan and goal drift management
Agents will drift. Plans will change. Decisions will get made. Are they being tracked?

A strong and truthful verifier
Use TDD and EDD, a solid continuous delivery mechanism, and a contained environment.

Human review layer
Build a rendering skill that gives us (humans) the right amount of information, without AI slop. Lowers review burden. Faster decisions and feedback for agent.

A word of caution: long-running agents != desired outcomes.

If you don’t get the factory design right, your long-running agents will deliver something. Not necessarily what you want.

[Dan Adler @DanielNealAdler]

Every. single. enterprise company I talk to is building (or already has) one of these right now

[Superflux AI @SuperfluxSF]

I want to know too

[Hermes Release Watch @HermesWatcher]

How many distinct Hermes profiles are you actively using?

One profile per project, role, channel, or recurring workflow.

[Superflux AI @SuperfluxSF]

A lot of problems with coding agents stem from ambiguous user messages (esp. when you are in the thick of iterating with the agent)

I created an ambiguity-filter plugin to make the agent catch input that feels ambiguous to itself by reflecting on the assumptions made in interpreting the user intent.

It has been a gamechanger for me.

I am testing and securing the plugin and will be release it soon for free.

If you are interested in beta testing this plugin, DM or comment .

[Teknium 🪽 @Teknium]

Hermes Agent now scales to a near infinite number of MCP and plugin tools with high accuracy. You save a ton of money on input tokens and lose almost no accuracy in our internal testing.

We've implemented progressive disclosure for tools, so they behave much more like skills. All MCP tools now sit behind a single search-and-execute tool.

When your MCP tools use little context, their names and descriptions are shown to the model directly. If that would use more than 5% of your available context, the agent is instead told it has tools from X MCP behind tool search, so it always knows they're available.

[𒐪 @SHL0MS]

just tell it no skeuomorphic metaphorical substitutions, or better yet use the Hermes Humanizer skill

[Teknium 🪽 @Teknium]

Just gave Hermes' skills an overhaul - all the built in and optional skills we support with some qol improvements, enhancements, reorgs, cleanup, and more.

Enjoy!

[Superflux AI @SuperfluxSF]

@dexhorthy A lot of good stuff in part 1 and part 2. Highly recommended reading

[dex @dexhorthy]

part one of "Why Software Factories Fail" hit 250k views in under 24h - now you get to enjoy part two https://t.co/NLkDvSTDhW

[Superflux AI @SuperfluxSF]

@aiedge_ Both of these are on by default → Persistent memory → Skill generation

[Teknium 🪽 @Teknium]

As much as i love automation with Hermes and ai agents, i agree- post your own slop yourself. If someone wanted llama 5b’s responses to their thoughts, theyd ask it themselves.

[Superflux AI @SuperfluxSF]

Build the assembly line so that anyone can build the cars. Safely, repeatably and correctly. https://t.co/mewoNyVqIP

[Jensen Huang @JensenHuang]

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.

The world needs both frontier closed models and frontier open models.

https://t.co/AUKzoQ5Ikb

[Superflux AI @SuperfluxSF]

https://t.co/ajyzNEAcPY