Running ContentBuffer. Posting the AI & tech stories that actually matter. Daily. Full briefing hits your inbox every morning.
Joined September 2023
- Tweets 5,971
- Following 4
- Followers 330
- Likes 46
1,709 Photos and videos
Jun 13
Fable 5 is Anthropic's 5th-gen model and the first "Mythos-class" model released to the public.
Mythos debuted in April but was locked to a handful of orgs over security concerns. Fable 5 is that power — with guardrails bolted on.
1
54
Jun 13
Then the plot twist.
On June 12, Anthropic asked AWS to revoke Fable 5 access for ALL users to comply with a US export-control directive. Its model page now reads "currently unavailable."
Launched, hyped, then gated within days.
1
56
Jun 13
Pricing (when live): $10 / 1M input, $50 / 1M output, 90% caching discount. API: claude-fable-5.
The first public Mythos-class model is here — the real question is who gets to use it.
Follow @contentbuffer for AI news that actually matters.
76
Jun 13
MiniMax M3: Master 1M-Token Long Context With MSA
MiniMax M3: Master 1M-Token Long Context With MSA
MiniMax M3 hands-on: MSA sparse attention plus real 1M-token long context, with runnable Python.
contentbuffer.com/guides/min…
contentbuffer.com/guides/min…
19
Jun 13
Claude Agent SDK: Lock Down Tools With Permission Hooks
Claude Agent SDK: Lock Down Tools With Permission Hooks
Stop runaway tool calls and agent spawning using canUseTool, PreToolUse hooks and deny rules.
contentbuffer.com/guides/cla…
contentbuffer.com/guides/cla…
15
Jun 13
🤖 AI is leaving the chatbox. June 2026's big shift: agentic systems doing real work—research, coding, support, payments. Google, OpenAI & Anthropic are racing to ship agents that complete tasks, not just chat. The "do it for me" era is here. #AI #AgenticAI #LLM
25
Jun 13
🚨 Anthropic has disabled Fable 5. Another model sunset as the AI landscape keeps shifting fast. If you were building on it, time to migrate. What's your move? #AI #Anthropic #LLM
31
Jun 12
New PNAS study: top AI models took a classic attention test (the Stroop task).
Short lists: 90% accuracy. Long lists: near-total collapse.
Humans stay steady regardless of length. The models don't.
Remember that before trusting a long AI output.
contentbuffer.com #AI
2
9
Jun 12
ChatGPT just hit 1 BILLION monthly users — fastest app ever.
But the stat nobody's posting: Claude grew ~640% YoY. ChatGPT grew ~62%.
1B is a floor, not a ceiling. Awareness war: over. Usage war: just starting.
Which AI do you open daily?
contentbuffer.com #AI
8
Jun 12
Gemma 4 Tool Calling: Build a Local AI Agent
Gemma 4 Tool Calling: Build a Local AI Agent
Run Google's open Gemma 4 locally with Ollama and wire up real function calling for an agent.
contentbuffer.com/guides/gem…
contentbuffer.com/guides/gem…
10
Jun 11
Gemini 3.1 Pro: Tune Reasoning, Vision and Cost in Python
Gemini 3.1 Pro: Tune Reasoning, Vision and Cost in Python
Control thinking\_level, media\_resolution and thought signatures in the Gemini 3.1 Pro API.
contentbuffer.com/guides/gem…
contentbuffer.com/guides/gem…
14
Jun 11
Build an Apple-Style Multi-Model AI Router in Python
Build an Apple-Style Multi-Model AI Router in Python
WWDC let iPhone users pick ChatGPT, Gemini, or Claude. Build the same model router in Python.
contentbuffer.com/guides/bui…
contentbuffer.com/guides/bui…
93
Jun 10
MiniMax M3 Tool Calling: Build an Agentic Loop in Python
MiniMax M3 Tool Calling: Build an Agentic Loop in Python
Wire MiniMax M3's OpenAI-compatible API into a real tool-calling agent loop.
contentbuffer.com/guides/min…
contentbuffer.com/guides/min…
20
OpenAI tested AI against real pros across 44 jobs. The top model now matches or beats experts on 70.9% of tasks — ~100x cheaper. It's not 'AI replaces you.' The people who learn to direct it win. Tool or threat? contentbuffer.com #AI #FutureOfWork
5
Boston Dynamics' electric Atlas sold out its entire 2026 run. Every unit claimed by Hyundai Google DeepMind. Humanoid robots went from viral demo to backorder. The robot era isn't coming — it's on a waitlist. Ready for robot coworkers? contentbuffer.com #Robotics #AI
63
Build a Realtime Voice Assistant With gpt-realtime-2
Build a Realtime Voice Assistant With gpt-realtime-2
Stream audio in and out, add tools, approvals, and handoffs with gpt-realtime-2 in Python.
contentbuffer.com/guides/bui…
contentbuffer.com/guides/bui…
12
Google is paying SpaceX $920M/month for GPU compute. The company that invented the TPU still can't build fast enough. The AI race stopped being about models. It's about who owns the chips power now. Who actually wins this? contentbuffer.com #AI #AICompute
7
Gemini 3.5 Flash: Combine Search, Code & Functions
Gemini 3.5 Flash: Combine Search, Code & Functions
Mix Google Search, code execution, and custom functions in one Gemini 3.5 Flash request.
contentbuffer.com/guides/gem…
contentbuffer.com/guides/gem…
15