David Kendrick - Practical AI

Building practical AI systems in public-agents, useful apps, local models, and lessons between prototype and reality.

[David Kendrick - Practical AI @DavidKendrickAI]

I'm David. I build practical AI systems—agents, automations, local-model experiments, browser workflows, and useful apps.I'll share what works, what breaks, and what I learn between a promising prototype and something people can actually rely on.https://t.co/sIbac3Mhd5

[clem 🤗 @ClementDelangue]

In the spirit of transparency, here’s what I asked @OpenAI:

• Radical transparency: let’s release the traces from the “rogue” agents so the entire research community can study what happened.

• More capabilities for defenders: let’s commit $100M in compute from OAI to help the Hugging Face community build powerful cyber defenses with the best open and closed models.

The first autonomous agent cyberattack is an unprecedented event. It deserves an unprecedented response!

[beamnxw ./ @beamnxw]

https://t.co/Qd15VI1bFb

[Shensi Ding @shensi]

We just beat the best frontier models at a quarter of the price. Now you can too, in one API call.

Today we're launching Merge Fusion.

Send one prompt to a panel of models, and a judge writes one answer better than any of them alone.

Fusion beats Fable 5 outright at 1/4th the cost. It also sets the highest score on the DRACO benchmark.

[Vox @Voxyz_ai]

𝘀𝘁𝗼𝗽 installing every Skill you come across in Claude Code/Codex.
𝘀𝘁𝗼𝗽 letting five Skills do the same job.
𝘀𝘁𝗼𝗽 keeping every Plugin and MCP enabled just in case.

you never built a 𝘀𝗸𝗶𝗹𝗹 𝗶𝗻𝘃𝗲𝗻𝘁𝗼𝗿𝘆. symlinks, duplicates and plugins blur together.

i checked mine:
→ 𝟭𝟱𝟰 SKILL.md files across 3 local directories
→ 43 symlinks
→ 34 identical-content groups
→ 21 enabled plugins

i couldn't tell which ones i actually used. some old Skills repeat work stronger models can do themselves.

Codex caps its initial Skills list at 𝟮% of context, or 8,000 characters if the window size is unknown. after that, it trims descriptions, then leaves Skills out.

i built a safe cleanup flow: read-only audit, offline HTML report, human approval before any merge, archive or disable, plus original paths and rollback steps.

7 prompts you can copy below 👇

[Tran Mau Tri Tam ✪ @tranmautritam]

Where I go when I run out of design ideas. Save this 👇

- https://t.co/MxeRSvBDxr - best of social post design
- https://t.co/5SM6zQRHrH - daily web design inspiration
- https://t.co/21RG68ReVa - minimal websites directory
- https://t.co/c7NA6ro0zG - OG images done right
- https://t.co/D9m4nLo37C - beautifully designed decks
- https://t.co/zX9pbvqced - 1,200+ logos by top designers
- https://t.co/ru7zu4VJhD - branding & editorial projects
- https://t.co/O58AD1CIyO - real brand guidelines library
- https://t.co/OPC3oYmGIe - best of recent design, daily

What am I missing?

[Celeris Labs @Celeris_ai]

Introducing celeris-1.

A general purpose language model which delivers near-GPT-5 level intelligence with 15x faster response times.

We’ve developed a new inference architecture that uses diffusion techniques instead of conventional autoregressive generation - unlocking dramatically better speed while maintaining frontier-level intelligence.

Celeris-1 delivers a p50 response latency of 157ms - around 15× faster than GPT-5-mini and 17× faster than GPT-5 - while scoring comparably 76% on MMLU-Pro, compared with 78% and 81%, respectively.

On tokens per second, Celeris-1 achieves a throughput of 1,280 tokens per second vs 144 for gemini-3.5 flash-light using a reconstructed version of @ArtificialAnlys's tokens per second benchmark dataset.

The model is available starting today. Sign up at https://t.co/99ywRD88KS

[Alex Kotliarskyi 🇺🇦 @alex_frantic]

How to graph-max with Codex and 5.6 Sol:

1. Draw a graph (literally in any tool, even on paper)
2. Send it to Codex and say "write a code mode script that implements this workflow, run it with <your inputs>"

There's no step 3, it just works.

Seriously it's that simple.

[Superlinear @superlinear_fm]

Understand the Loops and Graphs that Frontier Models Like Opus 5, Fable 5, and GPT-5.6 Use

In our new episode:
→ How agent loops become graphs
→ Why focused subagents can perform better
→ Where the extra tokens go
→ When parallelism helps
→ When orchestration becomes expensive overhead
→ How to diagnose a run that gets slow, costly, or stuck

You don’t need to design every graph yourself.

But you should understand the graph your harness is already building.

00:00 Introduction
03:34 How the Agent Loop Works
06:18 Goal Loops and Autoresearch Loops
14:57 From Loops to Graphs
17:06 Sub-Agents: Benefits, Costs, and Context
21:18 The Research Fan-Out Graph
26:27 Multi-Model Agent Graphs
28:37 Asynchronous Implementation and Review
33:40 How to Design Agent Graphs
36:07 When the Harness Builds the Graph for You: Token Spend vs. Performance
38:38 Execution Traces: A Profiler for Agents
41:46 Final Takeaway

[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!

[Meng To @MengTo]

I'm open-sourcing my Three.js game dev skills.

Build an isometric action RPG with camera controls, VFX, audio, monster assets, combat systems, and more.

Everything is free:
https://t.co/BoiGY8zvA9

Play the game:
https://t.co/ylWrybZMns

Game assets:
https://t.co/FlOc22IT5v

[David Kendrick - Practical AI @DavidKendrickAI]

I’ve learned that telling an AI agent to “make this better” rarely helps. I get better results when I say what’s wrong, ask it to change only what’s needed, and give it a few clear checks to pass.

[Alex Finn @AlexFinn]

Today’s the day.

Opus 5 and a super-fast version of ChatGPT 5.6 are dropping imminently

In fact, if you use Opus 4.8 right now, there’s a decent chance your prompts are getting routed to Opus 5

Days like this don’t happen often. It’s absolutely critical you lock in:

1. Test these models the moment they drop
2. If you don’t have your own personal benchmark tests based on your every day workflows and tasks, build them now (will open source mine soon)
3. Put these new models on your latest projects. See how they hold up against what you were doing before
4. Plug them into your Hermes/OpenClaw. Measure cost per task vs speed
5. Have these newer models review your code built by older models. Find security gaps/optimization opportunities the less intelligent models couldn’t find before

99.99% of the world won’t even know what’s happening today. You’re going to be the .01% that jumps on this and attacks

Lock in.

[Chaofan Shou @Fried_rice]

Kimi K3 exploited the latest Redis server with a 0day it discovered.

All it took was 27min with 32 agents.

https://t.co/ZOKA2wchLa

[Nous Research @NousResearch]

Hermes now has an optional skill for tldraw offline, @tldraw's desktop whiteboard.

It reads your open canvas, edits it live, and can embed a script in the file so the drawing opens with working buttons and state.

> hermes skills install optional-skills/creative/tldraw-offline

[Femke Plantinga @femke_plantinga]

A few weeks ago everyone was talking about loops. Now it's graphs.

Both live or die on one thing: your company brain.

This week, Shann Holmberg shared how he runs his marketing on graphs.

Here's the difference (resolving a support ticket):

𝗟𝗼𝗼𝗽𝘀 → you set the frame, the agent owns the path.
Hand it a goal and a bar. It drafts, checks, fixes, and loops until it clears.
You don't pick the steps. The agent does.

𝗚𝗿𝗮𝗽𝗵𝘀 → you draw the path first, the agent fills each node.
read → triage → draft → QA → send, with a checkpoint at each step.
The agent solves each box, but the map is yours.

Every one of those nodes pulls from the same place.

Triage needs past tickets. Draft needs the policy and product docs. That shared context (your company brain) is what each box reaches into as the work moves through. The brain has that map, the graph is that process, inferred explicitly.

𝗦𝗼 𝘄𝗵𝗲𝗻 𝗱𝗼 𝘆𝗼𝘂 𝘂𝘀𝗲 𝘄𝗵𝗶𝗰𝗵?
→ Reach for a 𝗹𝗼𝗼𝗽 when it's one-off work, no clear path yet. Let the agent figure it out.
→ Reach for a 𝗴𝗿𝗮𝗽𝗵 when it's repeatable work you already know the steps for. Lock them in.

Loops are for exploring. Graphs are for scaling.

Which one are you running right now? 👀

[slash1s @slash1sol]

A TEAM JUST DEPLOYED 15 AUTONOMOUS LOOP AGENTS FROM A SINGLE PROMPT USING APPLIED GRAPH ENGINEERING

Most developers still manually hardcode multi agent systems, writing separate logic for every individual task.

Graph engineering changes this by using a central topological map to spin up all 15 nodes simultaneously.

A single 200 word input generates the architecture, routing 120 unique pathways between agents instantly.

Instead of failing under conflicting instructions, these loop agents self correct via continuous state sharing.

Managing a 15 node mesh requires high token throughput, making this dependent on strict low-] latency API tiers.

See exactly how this automated multi agent graph architecture actually operates in real time ↓

[wast3 @0xWast3]

https://t.co/j9thHBFgdB

[Codez @0xCodez]

https://t.co/WVztfCUH4u

[Carlos E. Perez @IntuitMachine]

https://t.co/CHEk211f7Y

[divyansh tiwari @DivyanshT91162]

NVIDIA just dropped something insane.

They're giving free API access to 80+ AI models.

Including:

→ MiniMax M2.7 → GLM 5.1 → Kimi 2.5 → DeepSeek 3.2 → GPT-OSS-120B

And yes, it works with:

→ Cursor → Hermes → OpenCode

Setup:

• Get an API key from https://t.co/QwpfkOaOVT • Base URL: https://t.co/G5dfsJhsTv • Add your NVIDIA_API_KEY • Pick any model and start building.

80+ models.

$0 to get started.

NVIDIA is making it ridiculously easy for developers to experiment with frontier AI.