Tim Dentry

@mattjay I asked this same question

Everyone using Claude code and/or Codex - how are you enforcing them to not pull in new/potentially malicious packages from npm or PyPi?

@atmoio I love you @atmoio ! Your content just keeps getting better and better!

AI is not for you

This is fascinating. The perception of the race to the bottom for token costs is a complete myth (if you want your AI to be some kind of "good"). Some really good nuggets in Mo's discussion.

@github I wonder how much of the installation of poisoned extensions is being driven by Cursor, Codex and Claude Code. To my knowledge, none of the vendors are performing SCA for the supply chain that their models "recommend" when generating code.

1/ We are sharing additional details regarding our investigation into unauthorized access to GitHub's internal repositories. Yesterday we detected and contained a compromise of an employee device involving a poisoned VS Code extension. We removed the malicious extension version, isolated the endpoint, and began incident response immediately.

@H4ckmanac If you’re going to post this type of PSA, it would be nice if your details and impact assessment link was not to a marketing page.

🚨Data Breach Alert ‼️ 𝗧𝗲𝗮𝗺𝗣𝗖𝗣 𝗖𝗹𝗮𝗶𝗺𝘀 𝗦𝗮𝗹𝗲 𝗼𝗳 𝗚𝗶𝘁𝗛𝘂𝗯 𝗜𝗻𝘁𝗲𝗿𝗻𝗮𝗹 𝗦𝗼𝘂𝗿𝗰𝗲 𝗖𝗼𝗱𝗲 TeamPCP hacking group claimed the compromise and sale of GitHub internal data, allegedly including around 4,000 private repositories containing source code related to GitHub’s main platform and internal organizations. Threat actor: TeamPCP Sector: ICT Data exposure (claimed): Approximately 4,000 private repositories Data type: Source code Observed: May 19, 2026 Status: Pending verification ESIX©: 7.96 Full details and impact assessment on HackRisk.io

This one never gets old.

@constants2026 yay... the downfall of creativity. Why celebrate?

Claude can now REALLY generate songs for you.

Anthropic truly believes in the aramaic linguistic origin of "Abra Cadabra" (avra kedabra), which translates roughly to "I create as I speak" or "it came to pass as I spoke". This is exactly what Cisco did in the late 90s / early 2000s. They saturated the market with this concept of CCNAs, created a massive market of resellers (with various value levels) and effectively cornered the network plumbing market. To be fair, it's not a bad strategy.

The AI industry just invented a new job. Wait until you hear what it does.

@atmoio Didn't they used to call these folks "Sales Engineers" or "Solutions Engineers"?

This is great advice for engineers and mathematicians, simply reinforcing concepts that have been in existence since the beginning of the human race: "Excellence is never an accident. It is always the result of high intention, sincere effort, and intelligent execution." - from Aristotle. More contemporaneously from Greg Reid: “A dream written down with a date becomes a goal. A goal broken down into steps becomes a plan. A plan backed by action makes your dreams come true.”

A few major use cases for agentic coding for me: 1. Adhoc data visualizations. Anytime I have a question that can be answered quantitatively, I generate some code to make a plot. 2. Adhoc data annotation UIs. In ML, "make your own dataset" is often the answer, and that used to take a lot of custom UI work. 3. Adhoc CLIs for existing code. With visual elements.

Masterful remix. I wish that these remix experts would upload to Spotify.

@adamscochran @coinbase Karma….

The day after the CEO lays off a ton of staff and says: “Non-technical teams are now pushing code to production with AI” @coinbase has a major outage on their trading engine, and even their status page doesn’t work. 😂

#Hiring AI Engineers (3–5 yrs) Looking for builders with strong Python + GenAI experience. Core skills: • RAG pipelines • LLM integrations • LangChain / LangGraph • Vector DBs • Cloud deployment Bonus points for exposure to: • MCP • Multi-agent orchestration • AI memory systems • LLM fine-tuning • AWS Bedrock Fully remote role. We care more about shipped AI products than keyword matching. Please send in your application/resume to ng@askng.in

@RoundtableSpace How did you benchmark?

The best LLM setup there is right now: Frontend: Opus 4.7 Backend: GPT-5.5 Debugging: Sonnet 4.6 Research: Grok 4.2 Writing: GPT-5 Mini Image: Gemini 3.1 This is all you need.

@bindureddy Is this going to be Klarna v.2?

LITERALLY A RECIPE FOR DISASTER "Non technical teams shipping code to production" The LLMs - even the top ones - make a lot of mistakes and sometimes do very poor design

@MilkRoadAI I suspect these claims will be sliced and diced by the academic community and the frontier model vendors. Would be great if all of this was true, but unfortunately I am doubtful.

This is one of the craziest AI launches of 2026 and it came out of basically nowhere (Save this). A company called Subquadratic just shipped SubQ, and the benchmarks are almost hard to believe. To understand why this is such a big deal, you have to understand the fundamental problem that has defined AI for the last decade. Every large language model in existence is built on transformer architecture, and transformers use a mechanism called standard attention that checks every single word in a sequence against every other word. Double the context length and compute doesn't double, it quadruples, triple it and compute goes up nine times. This quadratic scaling is why frontier models have been stuck at roughly 1 million tokens, why running them at those lengths gets expensive fast, and why the AI labs have essentially been printing money charging you more the longer you need the model to think. The industry has known this problem existed since 2017 but they scaled it anyway. SubQ is built from the ground up to solve it. Instead of processing every possible token relationship, SubQ's sparse attention architecture identifies which relationships actually matter and ignores the rest meaning compute is used where it counts and wasted nowhere else. The result is that compute scales linearly with context length instead of exponentially, and the implications of that one architectural shift are enormous. At 12 million tokens, SubQ reduces attention compute by nearly 1,000x compared to standard frontier models and at 1 million tokens, it runs 52x faster than FlashAttention. And it does all of this while posting frontier level accuracy, scoring 95% on the RULER 128K long-context benchmark versus Claude Opus 4.6's 94.8%, and an 81.8 on SWE-Bench Verified coding tasks, besting Opus 4.6 (80.8) and DeepSeek 4.0 Pro. The cost comparison is where it gets genuinely insane. SubQ runs at under $1.50 per million tokens less than 5% of what Claude Opus charges. On the RULER benchmark, running the test with SubQ cost $8, running the same test with Claude Opus cost $2,600 and that's a 300x cost reduction at equivalent or better accuracy.. Subquadratic launched with $29 million in funding, SubQ is available today for early access via API, and SubQ Code, a coding agent built on the architecture ships alongside it. The transformer has been the unchallenged foundation of every major AI system since 2017. SubQ is the first serious evidence that something structurally better might have just arrived.

@TheEthanDing Remember, they are deeply in debt and racking up charges on the daily :-)

i have yet to meet a single person who feels like claude code is getting exponentially better on some kind of fast take off

@davieball Why would I do this?

Today we’re launching instructions.bio — portable biological instructions for AI. Connect them to ChatGPT, Claude, or any AI with memory, and every conversation becomes tailored to your biology.

That's an interesting point, but I think this would have to be viewed similar to solar. No city planner will EVER be "forced" to allow this requirement, or levy this "forced" requirement on their constituents. Having said that - i could see how, municipality/industry partnerships could take form and thus create incentives for constituents. So I think mandatory might be a little too pessimistic.

@Ja4h3ad @exec_sum They won't have a choice.... Building developers will make it mandatory if you purchase a house in their subdivision.... Cuz they'll get kick backs.

BREAKING: Nvidia and PulteGroup are partnering with startup Span to install mini data centers on the walls of new homes Each unit packs 16 Nvidia Blackwell GPUs, 4 AMD EPYC CPUs, and 3TB of RAM - and taps unused home electrical capacity to run AI inference workloads