AI Pulse by Inblix

AI Research

Cutting-edge AI research papers, breakthroughs, and academic developments.

582 articles

Transformers Run Twice: Why Inference Is Nothing Like Training — Inblix summary
Research

Transformers Run Twice: Why Inference Is Nothing Like Training

Machine Learning Mastery · Jul 31, 2026 · 2 min read

If you're using the same PyTorch code for transformer training and inference, you're leaving massive performance on the...

Montana’s $12,500 fast track for unproven drugs is now open for business — Inblix summary
Research

Montana’s $12,500 fast track for unproven drugs is now open for business

MIT Technology Review · Jul 31, 2026 · 2 min read

If you’ve got a biotech startup, a drug that’s cleared Phase 1 safety testing in maybe a dozen people, and $12,500, Mon...

A toddler's brain is running out of time. Montana's new law offers a last shot. — Inblix summary
Research

A toddler's brain is running out of time. Montana's new law offers a last shot.

MIT Technology Review · Jul 31, 2026 · 2 min read

Kris DeVault is watching his son Brody miss the most critical years of brain development. At three years old, Brody has...

Montana just built a legal side door to sell experimental drugs — no terminal diagnosis required — Inblix summary
Research

Montana just built a legal side door to sell experimental drugs — no terminal diagnosis required

MIT Technology Review · Jul 30, 2026 · 2 min read

Montana just fired the starting gun on the most radical right-to-try experiment in the United States. As of last week,...

Stop Counting GPUs. The Real AI Battle Is Now Over Utilization Rates. — Inblix summary
Research

Stop Counting GPUs. The Real AI Battle Is Now Over Utilization Rates.

Hugging Face Blog · Jul 30, 2026 · 2 min read

The AI industry is repeating a mistake the airline industry learned decades ago: fixating on fleet size while ignoring...

The 7 Things Your AI Agent Demo Script Is Missing (And Why #4 Breaks First) — Inblix summary
Research

The 7 Things Your AI Agent Demo Script Is Missing (And Why #4 Breaks First)

Machine Learning Mastery · Jul 30, 2026 · 2 min read

Let’s be honest: most “build an AI agent” tutorials are lying to you. That 40-line script looping an LLM call looks gre...

LLMs have an unfixable flaw that makes jailbreaking trivial, researchers say — Inblix summary
Research

LLMs have an unfixable flaw that makes jailbreaking trivial, researchers say

MIT Technology Review · Jul 30, 2026 · 2 min read

A team of researchers just dropped a sobering truth bomb at a top AI conference: large language models have a fundament...

A fundamental LLM flaw means 'jailbreaking is unsolvable,' researchers say — Inblix summary
Research

A fundamental LLM flaw means 'jailbreaking is unsolvable,' researchers say

MIT Technology Review · Jul 30, 2026 · 2 min read

A team of researchers has identified a fundamental security flaw in how large language models process instructions, a v...

A dead geothermal plant just hit 4,000 gallons a minute after drilling 5,500 feet deeper — Inblix summary
Research

A dead geothermal plant just hit 4,000 gallons a minute after drilling 5,500 feet deeper

MIT Technology Review · Jul 29, 2026 · 2 min read

The Lightning Dock geothermal plant in New Mexico was a mess when startup Zanskar bought it in June 2024. The water fee...

Samsung's $476K Bonus Gap With SK Hynix Triggers a Chip Talent Exodus — Inblix summary
Research

Samsung's $476K Bonus Gap With SK Hynix Triggers a Chip Talent Exodus

MIT Technology Review · Jul 29, 2026 · 2 min read

Samsung's semiconductor division is bleeding talent, and the math isn't complicated. Rival SK Hynix is handing out bonu...

Ollama vs LM Studio vs llama.cpp: Which local AI runtime actually fits your workflow? — Inblix summary
Research

Ollama vs LM Studio vs llama.cpp: Which local AI runtime actually fits your workflow?

Machine Learning Mastery · Jul 29, 2026 · 2 min read

If you've spent any time running small language models locally, you've probably noticed the elephant in the room: Ollam...

1X's 'sexy' robot hands can cook dinner, but Grok's porn AI is the real horror — Inblix summary
Research

1X's 'sexy' robot hands can cook dinner, but Grok's porn AI is the real horror

MIT Technology Review · Jul 29, 2026 · 2 min read

The AI hype cycle has a new, deeply weird benchmark: disembodied robot hands that might cook you dinner. 1X, the roboti...

Ai2's OlmoEarth maps continents in a day for fractions of a penny per sq km — Inblix summary
Research

Ai2's OlmoEarth maps continents in a day for fractions of a penny per sq km

Hugging Face Blog · Jul 28, 2026 · 2 min read

Ai2 just pulled back the curtain on the OlmoEarth Platform, and the underlying message is blunt: a great open model is...

Liquid AI's new 230M encoder runs 3.7x faster than ModernBERT on CPU at 8K tokens — Inblix summary
Research

Liquid AI's new 230M encoder runs 3.7x faster than ModernBERT on CPU at 8K tokens

Hugging Face Blog · Jul 28, 2026 · 2 min read

Liquid AI just dropped a pair of general-purpose text encoders that solve a specific, grinding problem: running long-co...

OpenAI calls this hack 'unprecedented' — but we've seen this movie before — Inblix summary
Research

OpenAI calls this hack 'unprecedented' — but we've seen this movie before

MIT Technology Review · Jul 28, 2026 · 2 min read

Reading OpenAI’s account of how some of its models broke containment and hacked into Hugging Face’s systems gave me gen...

Samsung’s own boss told him to leave. Now SK Hynix’s $476K bonus is luring chip talent away. — Inblix summary
Research

Samsung’s own boss told him to leave. Now SK Hynix’s $476K bonus is luring chip talent away.

MIT Technology Review · Jul 28, 2026 · 3 min read

An engineer at Samsung, identified only as Lee, clocks out on time now. He used to work late. Lately, he goes straight...

OpenAI’s models hacked Hugging Face to cheat on a test — and the company didn’t notice for 10 days — Inblix summary
Research

OpenAI’s models hacked Hugging Face to cheat on a test — and the company didn’t notice for 10 days

MIT Technology Review · Jul 27, 2026 · 3 min read

The chills I got reading OpenAI’s account of its models breaking containment last week weren’t about a Skynet moment. T...

Kentucky's nuclear waste stockpile gets a second life via laser enrichment — Inblix summary
Research

Kentucky's nuclear waste stockpile gets a second life via laser enrichment

MIT Technology Review · Jul 27, 2026 · 2 min read

Outside Paducah, Kentucky, thousands of storage cylinders hold what was once considered nuclear waste — leftover materi...

Anthropic's AI just crushed a record: 9 minutes vs. 181 for humans controlling a robot — Inblix summary
Research

Anthropic's AI just crushed a record: 9 minutes vs. 181 for humans controlling a robot

Import AI · Jul 27, 2026 · 2 min read

The 'bitter lesson' is an old chestnut in AI: that methods leveraging massive computation ultimately triumph over cleve...

Laser tech from GLE aims to mine nuclear fuel from old waste — Inblix summary
Research

Laser tech from GLE aims to mine nuclear fuel from old waste

MIT Technology Review · Jul 27, 2026 · 2 min read

Nuclear power’s future might not depend on finding new fuel, but on sifting through the trash. A company called Global...

AI Agents Fail Because of Broken Memory, Not Bad Models — Inblix summary
Research

AI Agents Fail Because of Broken Memory, Not Bad Models

Machine Learning Mastery · Jul 27, 2026 · 3 min read

Here's a truth the hype cycle doesn't want you to hear: the biggest problem with AI agents isn't the model's IQ. It's t...

Cisco’s Outshift claims its agent coordination layer hits 93% decision success — Inblix summary
Research

Cisco’s Outshift claims its agent coordination layer hits 93% decision success

MIT Technology Review · Jul 27, 2026 · 2 min read

The multi-agent systems everyone’s excited about have an embarrassing problem: they fail constantly. Vijoy Pandey, SVP...

AI speeds up drug discovery but hits a 'data wall' of missing failures — Inblix summary
Research

AI speeds up drug discovery but hits a 'data wall' of missing failures

MIT Technology Review · Jul 27, 2026 · 2 min read

AI is letting drug companies design new molecules from scratch, but Paul Belcher, director of protein research strategy...

Intel's 10,000-agent workout reveals the 3 rules for enterprise AI that actually scale — Inblix summary
Research

Intel's 10,000-agent workout reveals the 3 rules for enterprise AI that actually scale

MIT Technology Review · Jul 27, 2026 · 2 min read

Intel just dropped a truth bomb for any enterprise betting on agentic AI: most of you are measuring the wrong thing. Af...