AlexAImaginator

🤖 Vibe coding with LLMs + prompt craft 🖼️ Curating: AI 📰 Tech ⚡ Security — Best of X. Follow 💙 ♻️

[AlexAImaginator @TraffAlex]

🖥️ Best Local LLMs for Consumer GPUs — llama.cpp Guide (June 2026)

What I actually run on consumer hardware right now. Every model below runs via llama.cpp with a simple one-liner — no Docker, no Python env, no cloud.

━━━ 8-16GB VRAM ━━━

🔹 Gemma 4-12B (Google)
• Smartest model in this size class — competes with stuff 2× bigger
• Unsloth's MTP GGUFs: 162 tok/s vs 52 tok/s normal (3× speedup)
• Minimum 8GB VRAM recommended for Q4_K_M quant
• GGUF → https://t.co/VWp818MB3D

🔹 LFM2.5-8B-A1B (LiquidAI)
• Hybrid MoE, only 1B active params — absurdly fast for its size
• Perfect for 8-12GB cards, MacBooks, or anyone on a tight budget
• GGUF → https://t.co/ZbOs4mXJDq

━━━ 16-32GB VRAM ━━━

🔹 Qwen3.6-27B (Qwen)
• Scored 1.00 on tool-efficiency benchmarks — best local agent available
• 40 deterministic tasks, 32k/128k context needle tests — all passed
• GGUF → https://t.co/n7K3sPvliE
• MTP version (faster) → https://t.co/gwdfnJTzcy

🔹 Qwopus3.6-27B-v2 (Jackrong)
• Best quantization of Qwen3.6-27B — topped 5 agent & coding benchmarks (1200 samples)
• If you're running Q4, this is the one to grab
• GGUF → https://t.co/tV1DFqXnOD
• MTP version → https://t.co/PMqz7V5ewv

🔹 Gemma 4-31B QAT (Google/Unsloth)
• QAT variant with MTP draft head: 76-125 tok/s (1.67× speedup)
• Excellent for multi-agent / subagent workflows
• GGUF → https://t.co/FgVsUX0YOB

🔹 Nex-N2-Mini (Nex AGI)
• Post-train of Qwen3.5-35B-A3B — MoE with only 3B active params
• Fits on 16GB+ VRAM, overflow loads from system RAM
• Adaptive thinking saves ~20% tokens with no quality loss
• For deep multi-step reasoning, nothing in this size comes close
• GGUF → https://t.co/oyC522a8Eh

━━━ Quick Picks ━━━

• 16GB all-rounder → Gemma 4-12B with MTP GGUFs
• 32GB all-rounder → Qwen3.6-27B / Qwopus-v2
• Agents & tool use → Qwen3.6-27B or Qwopus Q4
• Deep reasoning → Nex-N2-Mini (MoE, fits 16GB+)
• Tight budget → LFM2.5-8B-A1B
• Cheapest full build: 1× used RTX 3090 (24GB) + rest of PC ≈ $1000-1500

━━━ Setup on Windows ━━━

1. Download llama.cpp → https://t.co/et0J7Swua7 (latest .zip)
2. Extract to any folder (e.g. C:\llama.cpp)
3. Download a .gguf from the links above (Q4_K_M or Q5_K_M for best quality/speed balance)
4. Run one of the commands below depending on your hardware

━━━ Launch Commands ━━━

SINGLE GPU — Standard model (no MTP):

llama-server.exe ^
-m C:\models\Qwen3.6-27B-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
-ngl 100 ^
-np 1 ^
--port 8080 ^
--jinja

SINGLE GPU — MTP model (faster inference):

llama-server.exe ^
-m C:\models\Qwen3.6-27B-MTP-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
--spec-type draft-mtp ^
--spec-draft-n-max 3 ^
-ngl 100 ^
-np 1 ^
--port 8080 ^
--jinja

DUAL GPU — Split across two cards:

llama-server.exe ^
-m C:\models\Qwen3.6-27B-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
-ngl 100 ^
--tensor-split 0.55,0.45 ^
--main-gpu 0 ^
-np 1 ^
--port 8080 ^
--jinja

DUAL GPU + MTP + Vision (multimodal):

llama-server.exe ^
-m C:\models\Qwen3.6-27B-MTP-Q5_K_M.gguf ^
--ctx-size 180000 ^
--flash-attn on ^
--cache-type-k q4_0 ^
--cache-type-v q4_0 ^
--batch-size 1024 --ubatch-size 512 ^
--spec-type draft-mtp ^
--spec-draft-n-max 3 ^
-ngl 100 ^
--tensor-split 0.60,0.40 ^
--main-gpu 0 ^
-np 1 ^
--port 8080 ^
--jinja ^
--mmproj C:\models\mmproj-F16.gguf

━━━ Parameter Breakdown ━━━

-m <path>
Path to your .gguf model file. Change this to wherever you downloaded it.

--ctx-size 180000
Context window in tokens. 180k = huge context for long conversations or big codebases.
Reduce to 32768 or 65536 if you don't need long context — uses less VRAM.

--flash-attn on
Flash Attention — dramatically speeds up inference and reduces VRAM usage.
Works on RTX 30xx/40xx/50xx. Always enable this.

--cache-type-k q4_0 / --cache-type-v q4_0
Quantizes the KV cache (key/value attention cache) to 4-bit.
This is what makes 180k context fit in VRAM. Without it, huge contexts eat all your memory.
Quality impact is minimal — this is a free performance win.

--batch-size 1024 / --ubatch-size 512
batch-size = how many tokens are processed in one forward pass (throughput).
ubatch-size = micro-batch actually sent to the GPU per step.
Higher = faster prompt processing but needs more VRAM.
If you run out of VRAM, lower these (e.g. 512/256).

-ngl 100
Number of layers to offload to GPU. 100 = all layers on GPU (full offload).
This is what you want if the model fits in your VRAM.
If it doesn't fit, reduce this (e.g. -ngl 40) — remaining layers run on CPU/RAM.

--tensor-split 0.55,0.45
How to split model layers across multiple GPUs. Values are ratios.
0.55,0.45 = GPU 0 gets 55% of layers, GPU 1 gets 45%.
Adjust based on your VRAM — give more to the card with more memory.
Example: 0.70,0.30 for a 24GB + 12GB setup.
Not needed for single GPU setups.

--main-gpu 0
Which GPU handles the batch computation (the "orchestrator").
Set to 0 (your primary GPU). The other GPU(s) handle their assigned layers.
Minor performance impact — usually just leave it at 0.

-np 1
Number of parallel slots (concurrent requests). 1 = one user at a time.
Increase to 2-4 if you want multiple clients connected simultaneously.
Each extra slot uses additional VRAM for its own KV cache.

--port 8080
Which port the server listens on. Change if port 8080 is busy.

--jinja
Enables Jinja2 template processing — required for proper chat formatting.
Most modern models expect this. Always include it.

--spec-type draft-mtp
Enables Multi-Token Prediction (MTP) speculative decoding.
Only works with MTP GGUF models (downloaded separately).
The model predicts multiple tokens at once and verifies them — big speed boost.

--spec-draft-n-max 3
How many tokens the MTP draft head proposes per step.
3 is a good default. Higher = potentially faster but more VRAM and may reduce quality.

--mmproj <path>
Path to the multimodal projector file (for vision models).
Enables image understanding — paste screenshots into the web chat.
Only needed if you want vision capabilities. Omit for text-only use.

━━━ Your Hardware → Your Command ━━━

Single GPU (8-24GB VRAM):
Use the "Single GPU" command. Change -m to your model path.
8GB card → Gemma 4-12B Q4 or LFM2.5-8B
12GB card → Gemma 4-12B Q5/Q6
16GB card → Gemma 4-31B QAT Q4 or Nex-N2-Mini
24GB card → Qwen3.6-27B Q4/Q5, Qwopus-v2, Gemma 4-31B QAT Q5/Q6

Dual GPU:
Use the "Dual GPU" command. Adjust --tensor-split based on your VRAM ratio.
24GB + 24GB → --tensor-split 0.50,0.50
24GB + 12GB → --tensor-split 0.70,0.30
24GB + 8GB → --tensor-split 0.75,0.25

Want speed? Use MTP versions of models with the "MTP" commands.

Want vision? Add --mmproj with the projector file from the model's HuggingFace repo.

5. Once running, you get:
• Web chat UI → http://localhost:8080
• OpenAI-compatible API → http://localhost:8080/v1
• Playground → http://localhost:8080/playground

━━━ Why /v1 API Is the Killer Feature ━━━

One local endpoint replaces your entire cloud API bill. The /v1 endpoint is drop-in OpenAI-spec compatible — every tool that speaks OpenAI just works. No custom code, no glue layer.

Works out of the box with:
• IDEs: Cursor, Continue, Windsurf, Cline, Roo Code
• CLI tools: aider, Open Interpreter, OpenCode
• Frameworks: LangChain, LlamaIndex, LiteLLM
• Any OpenAI SDK (Python, Node, Go, Rust)

Why this beats cloud APIs:
• 100% private — code never leaves your machine
• $0 per token — no rate limits, no quotas, no surprise bills
• Works fully offline
• Zero telemetry, no training on your data
• Swap models by dropping in a different .gguf — no app changes needed
• Run 32k–128k context windows without burning money

Good combos:
• Cursor + Qwopus-v2 → near-frontier quality, zero API cost
• Continue + Qwen3.6-27B → best local coding agent
• aider + Gemma 4-12B MTP → 162 tok/s, feels instant
• OpenCode + Nex-N2-Mini → deep reasoning on 16GB

Set any OpenAI-compatible client to your local endpoint:
set OPENAI_API_KEY=sk-dummy (any non-empty string works)
set OPENAI_BASE_URL=http://localhost:8080/v1
# every OpenAI-compatible tool now hits your local GPU

Shoutouts: @0xSero @rS_alonewolf @witcheer @UnslothAI @LottoLabs

[AlexAImaginator @TraffAlex]

That's today's radar. 📡

Who else ran a share thread today?
Name them below — tomorrow's list grows.

Don't like being featured? Just say so — we'll leave you off.

Follow @TraffAlex so you don't miss tomorrow's radar.

[AlexAImaginator @TraffAlex]

Art Radar 📡 — 12 Nov

7 creators opened their share threads today.
Every challenge + theme + link in this thread ↓

Drop your art in THEIR thread — or right here.
One thread, all the doors open. 🎨

[AlexAImaginator @TraffAlex]

7/7 @artbymorbid — "Show your art. You never know who's watching."
no theme — anything goes
→
https://t.co/R3naSPsqDQ

[morbidnumbskullart @artbymorbid]

Show your art. 🎨You never know who’s watching.

[AlexAImaginator @TraffAlex]

🔒 CYBERSECURITY, PRIVACY & OPEN SOURCE ROUNDUP — November 12, 2026

1️⃣ A SECURITY LAB SCORES 33 CVEs IN THE LINUX KERNEL WITH AI-ASSISTED AUDITS

Microsoft's FORGE Lab published an unusual scorecard for machine-assisted kernel auditing: 59 bug reports filed, 33 of them assigned CVEs, 38 fixes already merged upstream, and eight vulnerabilities with lab-confirmed local root exploits. The write-up walks through CVE-2026-64563, where a stale rhashtable pointer turns into a privilege escalation. It is the clearest evidence so far that automated auditing produces real, patchable kernel findings, which moves the bottleneck onto maintainer review time rather than bug discovery.
🔹 @TweetThreatNews

2️⃣ A BLANK AUTHENTICATION KEY PUT REMOTE MANAGEMENT TOOLS ON 288 HOSTS

Huntress tracked rogue ScreenConnect clients and the SCKit toolkit on 288 endpoints across 48 organisations, linking the activity to an authentication flaw in RMM+ versions 1.0.31 and 1.0.32 that accepted an empty key. The toolkit's persistence is documented in detail, while the initial access path and any successful lateral spread remain unconfirmed. Legitimate remote management is now one of the most abused access routes in intrusions, precisely because the traffic looks like the help desk.
🔹 @TweetThreatNews

3️⃣ A CI SECRET STEALER WAS PLANTED IN THOUSANDS OF REPOSITORIES

Aikido and SafeDep report a campaign that used stolen GitHub credentials to plant a stealer inside repository CI configuration, collecting workflow secrets and the cloud credentials those jobs depend on; where the poisoned job ran, credentials were likely harvested. Source control accounts are the highest-leverage supply-chain target in a modern build pipeline, because one reused login reaches every secret that signs a release.
🔹 @TweetThreatNews

4️⃣ A BROWSER EXTENSION BECAME THE COMMAND CHANNEL FOR ESPIONAGE CAMPAIGNS

Palo Alto's Unit 42 detailed Falcon, a browser-based command-and-control framework used against government entities in at least five countries. Its extensions gather credentials and communications, hijack active sessions, and act as a front for Antino, a Rust implant that routes its own control traffic through cloud storage. Exfiltration that travels inside traffic to a mainstream storage provider is exactly the kind of path a perimeter rule never learns to block.
🔹 @TweetThreatNews

5️⃣ AN ATTACKER USED AI AGENTS TO STEAL SECRETS AND BILL A ROUND 740,000 DOLLARS

Hosting platform Railway published a post-mortem describing how an attacker combined AI agents, a flaw in the build system and permissive cloud access to lift customer API keys and generate roughly 740,000 dollars in charges. The breakdown is instructive because nothing exotic happened: an automation layer simply made reconnaissance, credential hunting and repetition far faster than a human intruder could manage.
🔹 @The_Cyber_News

6️⃣ A RELAUNCHED GAME STORE REMOVED ITS SECOND LOGIN FACTOR

PrivacyAffairs reports that the relaunched Humble Bundle site no longer offers a second step at sign-in, presenting a password, an emailed code and a Google sign-in as alternatives instead of layers; the author verified password-only access, and the company has not publicly commented. A small regression with an old lesson attached: account login and recovery, not encryption, remain the softest door in most services.
🔹 @TweetThreatNews

7️⃣ OPEN SOURCE SUPPORT FOR AMD'S NEXT GPU GENERATION HAS ALREADY STARTED

Initial AMD RDNA5 and GFX13 code has been sent out for the Mesa RADV and RadeonSI drivers, so open source support for an unreleased hardware generation is beginning in parallel with the proprietary stack. Drivers landing before the hardware ships remains one of the most underrated advantages of the open source graphics stack: it is the reason a new card is usable on a Linux desktop on day one.
🔹 @phoronix

8️⃣ A NEW WAYLAND COMPOSITOR IS BEING BUILT TO KEEP MATE ALIVE

Polo debuted as a new Wayland compositor written for the MATE desktop, an attempt to keep a long-lived environment working as the X11 infrastructure underneath it keeps thinning out. Legacy projects face the same fork in the road everywhere: rewrite the foundation and preserve the workflow, or quietly fade away with the display server. A purpose-built compositor is the more ambitious answer, and the one that keeps the desktop rather than replacing it.
🔹 @phoronix

💭 What ties today together is that the workload of finding problems has been automated faster than the capacity to review the fixes. A machine-assisted pass over the kernel produced more CVEs than most maintainers can triage in a season, an attacker's agents turned a build system into a billable wallet, and one remote management tool accepting an empty key was enough to reach dozens of organisations. Discovery stopped being the scarce resource; attention, credential hygiene and patch windows are.

Which one gets fixed first in your environment: the kernel backlog nobody has time to review, or the remote access tool nobody re-audits? 👇

#Cybersecurity #OpenSource #InfoSec #Privacy #Linux

[AlexAImaginator @TraffAlex]

@jasondainter you're welcome !

[AlexAImaginator @TraffAlex]

🎨 AI ART & VISUALS ROUNDUP — November 12, 2026

1️⃣ SEEDANCE 2.5 IS NOW RUNNING AT 4K INSIDE KREA

Krea added the 4K build of Seedance 2.5 and is calling it the highest quality AI video model it has shipped to the browser suite so far. That sits on top of the Video 2.0 tooling the company pushed last week, meaning cinematic camera controls, presets, style transfer and generated sound around the same model instead of a separate app. A live walkthrough with Numan Khan runs today at 10 am Pacific, aimed at people already cutting finished footage inside the browser rather than exporting it elsewhere.
🔹 @krea_ai

2️⃣ ONE OPEN SOURCE NODE CUTS A MINI MAX H3 CLIP TO 72 SECONDS

A community node rewrote the MiniMax H3 generation path so a five second, 24 fps character clip renders in roughly 72 seconds warm and 95 seconds on a cold start on a single RTX 4090, against about 650 seconds natively. Peak video memory drops from around 22.5 GB to 13.5 GB, which is the difference between a flagship card and a mid range one, and it runs on a four step Euler schedule. Single node, open source, published as a one file drop in.
🔹 @Nyxelarm

3️⃣ FASTGEN OPENS DISTILLATION UP TO H3, LTX 2.5 AND FLUX.2

The FastGen library shipped a large update covering model support and acceleration methods in one pass. New models: H3, LTX 2.5, LingBot World, Cosmos 3 and FLUX.2. New methods: PDD, AYF, AnyFlow and Causal Forcing, plus multimodal distillation, image editing, camera conditioning and key value caching. For anyone running few step video or image models on their own hardware, distillation tooling that keeps pace with this many releases in one release is rarer than it should be.
🔹 @julberner

4️⃣ ULTRATEX PUTS READABLE TEXT ON 3D MODELS, LOCALLY AND OPEN

UltraTex went open source: a texturing model built on the FLUX.2 Klein 9B base with a dedicated LoRA, generating 2K textures for 3D meshes and handling clean, legible lettering, which is where most texture generators fall apart. It runs locally for free, weights are already released, and there is a ComfyUI workflow wired up for it. For game and product work that means asset texturing without a licensed pipeline or a render farm.
🔹 @philippsieben

5️⃣ WHAT VIDEO MODELS ACTUALLY READ OUT OF A 3D BLOCKOUT

ComfyUI published a breakdown of what generative video models such as Seedance 2.5 and MiniMax H3 pick up from a rough 3D blockout, working from crude masses and simple camera moves all the way through to a fully animated character. The value is the graded comparison: instead of one impressive clip, it shows which staging information survives the handover from blocking to generation and which gets ignored. That determines whether blocking a shot in 3D before generating is a real workflow step or just extra labour.
🔹 @ComfyUI

6️⃣ A FULL MUSIC VIDEO BUILT AT 99 PERCENT AI

Jason Dainter published a finished music video and put the pipeline in the caption: stills from Midjourney v8, motion from MiniMax H3, music from Suno v6, lyrics and cutting done together with Claude Opus 5.5, assembled in DaVinci Resolve. More than a quarter million views and heavy reposts, with the replies split between the tool list and whether authorship belongs to whoever wrote the prompts. It is one of the clearer published credits lists for a long form generative music video.
🔹 @jasondainter

7️⃣ THE MOST CONVINCING THIRTY SECONDS OF NOTHING YET

A 30 second clip of grass being mowed, generated with GPT 6, Seedance 2.5 and Midjourney plus light post processing, and the point is the boringness: no morphing blades, no camera drift, no dream logic, just a lawn. Hyperreal mundanity has quietly become the benchmark people use to test whether a video release is worth the credits, and mowing footage is the honest version of that test.
🔹 @Arvin010717

8️⃣ THE ARTIST WITH ZERO SKILLS IS NOW A 200 MILLION VIEW CASE STUDY

Linus Ekenstam took stock of Antrogfrog, the account that exploded over the past forty days and whose single animated short, made with Midjourney and Krea, ran somewhere around 200 million views in 36 hours across platforms. The framing is deliberately provocative, an artist with no traditional training producing the strongest storytelling piece in the feed, and it keeps landing because the audience reaction is the data, not the tool list.
🔹 @LinusEkenstam

💭 Two speeds in the same morning. On one side the plumbing is getting serious, a community node halving the wait for a 24 fps clip, a distillation library catching five model families at once, an open texturing model that can spell. On the other side the work people actually watch, a music video, a lawn, a short with two hundred million views, and none of it depends on which checkpoint was used. The gap between the pipeline threads and the art threads is where most of the argument about this medium is really happening.

Which one are you testing today, the 72 second H3 node, the FastGen distillation update, or a 4K Seedance clip of absolutely nothing? 👇

#AIArt #GenerativeAI #AIVideo

[AlexAImaginator @TraffAlex]

🤖 AI NEWS ROUNDUP — November 12, 2026

1️⃣ CHATGPT IS BECOMING A SHOPPING WINDOW FOR REAL BUSINESSES

OpenAI opened a beta of Businesses in ChatGPT: owners can claim their listing, manage the information people see, and get analytics on how customers engage with them inside the assistant. The bigger piece is the Business Agent — eligible companies can publish their own agent, clearly labelled and kept separate from ChatGPT's independent answers, so people can ask questions, compare offers and find what fits their needs. It is the clearest sign yet that assistant surfaces are being treated as distribution.
🔹 @OpenAI

2️⃣ GEMINI ROBOTICS 2.5 TEACHES ROBOTS EVERYDAY LABOUR

DeepMind's latest physical AI stack picks up hundreds of everyday skills instead of one scripted routine — folding laundry, serving pastries, and autonomously sorting car parts on a factory floor. The pitch is breadth and transfer: one model family that keeps adding tasks rather than a bespoke build per workstation.
🔹 @GoogleDeepMind

3️⃣ SCRIBE V3 WENT AFTER THE HARDEST PART OF VOICE: NOISE

ElevenLabs shipped Scribe v3 and Scribe v3 Live, its most accurate speech-to-text models, built for messy real-world audio. The model separates pauses from real sentence ends, tracks who said what in a crowded meeting, and the Live version adds diarization for the first time while cutting errors 43% versus its predecessor. Transcription is quietly turning into infrastructure for everything downstream.
🔹 @ElevenLabs

4️⃣ A MODEL THAT READS DNA ONE CELL AT A TIME

DeepMind introduced AlphaGenome 2, a regulatory genomics model that zooms into specific cell types instead of reading whole tissue as one blurry signal. Trained on more than 100M profiles across 68 tissues, it spots nearly twice as many disease-linked mutations — the kind of filter that turns a needle-in-a-haystack search for a rare disease cause into a shortlist.
🔹 @GoogleDeepMind

5️⃣ OPENAI IS PACKAGING ITS OWN SECURITY SCARS INTO PRODUCTS

After saying its frontier models found and resolved important vulnerabilities in OpenAI's own code and infrastructure, the company is turning that internal work into tooling and playbooks for customers, headlined by a live keynote on cyber at the frontier. Defender-grade access, validated fixes and internal hardening lessons are becoming a product line, not a blog post.
🔹 @OpenAI

6️⃣ AGENTS THAT ANSWER WITH SCREENS INSTEAD OF WALLS OF TEXT

Google released the first stable release of A2UI, an open protocol and framework for generative UI across web, Android, iOS, wearables and televisions. Instead of a paragraph, an agent can respond with the interface the task actually needs — a form to confirm, a chart to compare options. Standardised, machine-renderable interfaces may matter more than the model behind them.
🔹 @GoogleCloudTech

7️⃣ THE VOICE AGENT STACK JUST GREW MEMORY AND A BRAIN

At its New York summit, ElevenLabs previewed an orchestration layer called Interaction Engine that routes simple requests to cheap models and hard ones, like refunds, to frontier models; enterprise fine-tuning of its speech models and open-source language models on customer data; Routines that do the work before and after a call, from pre-screening an applicant to texting a customer about a confusing bill; Memory in its agents; and a workplace assistant in open beta. Cost control and follow-through, not voice quality, are now the battleground.
🔹 @ElevenLabs

8️⃣ ANOTHER SERIOUS OPEN-WEIGHTS AGENTIC MODEL LANDS UNDER A PERMISSIVE LICENCE

The Shanghai Innovation Institute released Nex-N2.6, a multimodal agentic model built on GLM-5.3-Flash, under the Apache 2.0 licence. The claims worth testing: 90% less repetitive reasoning, dropping from 16% to 1.6%, and full adherence to the tool list it is given. Less reasoning token burn on the same task is exactly where open models keep closing the gap.
🔹 @AdinaYakup

💭 Read together, this week's announcements barely mention benchmarks. Robots sorting parts, agents that remember and follow up, an interface protocol, a verification-heavy security product line, a genomics model that narrows a search space — the money is moving from raw model quality to the plumbing that makes model output usable, cheap and trustworthy at scale. The labs that own that plumbing will be the ones businesses cannot switch away from.

Which of these would actually change your work this quarter: agents that act on your calendar and calls, generative interfaces, or open models you can fine-tune on your own data? 👇

#AI #ArtificialIntelligence #MachineLearning

[AlexAImaginator @TraffAlex]

That's today's radar. 📡

Who else ran a share thread today?
Name them below — tomorrow's list grows.

Don't like being featured? Just say so — we'll leave you off.

Follow @TraffAlex so you don't miss tomorrow's radar.

[AlexAImaginator @TraffAlex]

Art Radar 📡 — 11 Nov

10 creators opened their share threads today.
Every challenge + theme + link in this thread ↓

Drop your art in THEIR thread — or right here.
One thread, all the doors open. 🎨

[AlexAImaginator @TraffAlex]

10/10 @Preda2005 — "Share your ART — I See YOU."
no theme — anything goes
→
https://t.co/SJIzE2Goww

[Marcio Lima 利真 マルシオ 💎 @Preda2005]

Share your ART 𓂀

I See YOU。