close

DEV Community

Breach Protocol profile picture

Breach Protocol

Plain-language AI news and curated, cited lessons — every claim verified against the original paper or the lab's own page. No aggregator hearsay, no AI slop.

Joined Joined on 
A 753 billion parameter model ran on a single workstation GPU
breachprotocol profile

A 753 billion parameter model ran on a single workstation GPU

#inference #localai #mixtureofexperts #efficiency
Comments
4 min read
A foreign-government contract paid for websites built to be quoted by chatbots
breachprotocol profile

A foreign-government contract paid for websites built to be quoted by chatbots

#cybersecurity #aisecurity #datapoisoning #supplychain
Comments
4 min read
A prompt injection that copies itself from agent to agent
breachprotocol profile

A prompt injection that copies itself from agent to agent

#cybersecurity #promptinjection #aisecurity #multiagentsystems
Comments
4 min read
Agent skills work by anchoring procedure, not by adding knowledge
breachprotocol profile

Agent skills work by anchoring procedure, not by adding knowledge

#agents #agentskills #research #contextengineering
Comments
4 min read
DeepSeek is selling a checkpoint it has not published
breachprotocol profile

DeepSeek is selling a checkpoint it has not published

#deepseek #openweights #pricing #inference
Comments
4 min read
Etched raised 700 million dollars and shipped its first rack to Jane Street
breachprotocol profile

Etched raised 700 million dollars and shipped its first rack to Jane Street

#hardware #inference #funding #etched
Comments
4 min read
Many AI-generated images cannot be traced to any training example
breachprotocol profile

Many AI-generated images cannot be traced to any training example

#diffusionmodels #copyright #provenance #interpretability
Comments
4 min read
Models change their behavior when they think a safety researcher is asking
breachprotocol profile

Models change their behavior when they think a safety researcher is asking

#alignment #evaluations #interpretability #anthropic
Comments
4 min read
One person with AI matched a two-person team at Procter and Gamble
breachprotocol profile

One person with AI matched a two-person team at Procter and Gamble

#labor #productivity #research #fieldexperiment
Comments
4 min read
Stripe is buying the company that keeps score on every model
breachprotocol profile

Stripe is buying the company that keeps score on every model

#industry #acquisitions #openrouter #stripe
Comments
4 min read
A discovery system that refuses to trust the language model's own confidence
breachprotocol profile

A discovery system that refuses to trust the language model's own confidence

#aiforscience #discovery #bayesianoptimization #drugdesign
Comments
4 min read
A runbook, not a model, hit 95 percent on a live agent benchmark for 15 dollars
breachprotocol profile

A runbook, not a model, hit 95 percent on a live agent benchmark for 15 dollars

#agents #harness #benchmarks #terminalbench
Comments
4 min read
A tool that strips SynthID and C2PA marks passed 4,900 stars and shipped again on August 18
breachprotocol profile

A tool that strips SynthID and C2PA marks passed 4,900 stars and shipped again on August 18

#cybersecurity #provenance #watermarking #c2pa
Comments
4 min read
AlphaEvolve helped tighten the matrix multiplication exponent, and the proof was checked in exact arithmetic
breachprotocol profile

AlphaEvolve helped tighten the matrix multiplication exponent, and the proof was checked in exact arithmetic

#mathematics #alphaevolve #deepmind #algorithms
Comments
4 min read
Claude designed protein binders against 14 of 15 targets and two labs built every one of them
breachprotocol profile

Claude designed protein binders against 14 of 15 targets and two labs built every one of them

#anthropic #proteindesign #aiforscience #drugdiscovery
Comments
4 min read
DFlash 2 decodes 3.4 times faster, and published the table showing where that stops being true
breachprotocol profile

DFlash 2 decodes 3.4 times faster, and published the table showing where that stops being true

#inference #speculativedecoding #efficiency #openweights
Comments
4 min read
Frontier multimodal models still cannot build a 3D world, and a new benchmark says under 60 percent
breachprotocol profile

Frontier multimodal models still cannot build a 3D world, and a new benchmark says under 60 percent

#multimodal #agents #3d #benchmarks
Comments
4 min read
OpenAI put its largest frontier training run on hold and priced the safety tax at 20 percent
breachprotocol profile

OpenAI put its largest frontier training run on hold and priced the safety tax at 20 percent

#cybersecurity #aisecurity #openai #frontiersafety
Comments
4 min read
Sainsbury's paused face scanning in one store after throwing out an innocent shopper
breachprotocol profile

Sainsbury's paused face scanning in one store after throwing out an innocent shopper

#facialrecognition #surveillance #biometrics #aigovernance
Comments
4 min read
Seven senators demand Apple reject Chinese memory chips as AI demand drains global supply
breachprotocol profile

Seven senators demand Apple reject Chinese memory chips as AI demand drains global supply

#policy #semiconductors #supplychain #memory
Comments
4 min read
Unitree lists in Shanghai as the rare profitable humanoid robot maker
breachprotocol profile

Unitree lists in Shanghai as the rare profitable humanoid robot maker

#robotics #humanoids #unitree #ipo
Comments
4 min read
An AI agent rewrote 189 files in a 717,000-line codebase and nobody read the code
breachprotocol profile

An AI agent rewrote 189 files in a 717,000-line codebase and nobody read the code

#codingagents #softwareengineering #casestudy #agents
Comments
4 min read
Anthropic still will not ship the model that found ten thousand vulnerabilities
breachprotocol profile

Anthropic still will not ship the model that found ten thousand vulnerabilities

#cybersecurity #aisecurity #vulnerabilities #redteaming
Comments
4 min read
Beijing's robot 100-meter dash goes fully autonomous this month
breachprotocol profile

Beijing's robot 100-meter dash goes fully autonomous this month

#robotics #humanoidrobots #embodiedai #china
Comments
4 min read
Deepfake detectors for crisis video collapse once the clip gets reshared
breachprotocol profile

Deepfake detectors for crisis video collapse once the clip gets reshared

#deepfakes #videogeneration #benchmarks #misinformation
Comments
5 min read
Cursor built its own GitHub
breachprotocol profile

Cursor built its own GitHub

#developertools #codingagents #productlaunch #cursor
Comments
4 min read
Frontier agents produced three genuinely new ideas out of 252 attempts
breachprotocol profile

Frontier agents produced three genuinely new ideas out of 252 attempts

#agents #evaluation #benchmarks #airesearch
Comments
4 min read
llama.cpp tagged v0.1.0 after ten thousand build numbers
breachprotocol profile

llama.cpp tagged v0.1.0 after ten thousand build numbers

#localinference #opensource #llamacpp #speculativedecoding
Comments
4 min read
Models are being trained to give up
breachprotocol profile

Models are being trained to give up

#reasoning #testtimecompute #efficiency #refusal
Comments
4 min read
The cheapest way to teach a model turned out to be blindfolding the student
breachprotocol profile

The cheapest way to teach a model turned out to be blindfolding the student

#distillation #trainingmethods #visionlanguage #efficiency
Comments
4 min read
The executive order people keep reading as a license to hack back
breachprotocol profile

The executive order people keep reading as a license to hack back

#cybersecurity #policy #aisecurity #law
Comments
4 min read
A world model that keeps the map outside the model
breachprotocol profile

A world model that keeps the map outside the model

#worldmodels #videogeneration #memory #interactive
Comments
4 min read
An AI scam agent got more people to comply than human operators did
breachprotocol profile

An AI scam agent got more people to comply than human operators did

#cybersecurity #aisecurity #socialengineering #fraud
Comments
4 min read
MiniMax put its video-with-sound model on Hugging Face
breachprotocol profile

MiniMax put its video-with-sound model on Hugging Face

#videogeneration #openweights #minimax #china
Comments
4 min read
OpenRouter now picks your model by what everyone else is paying for
breachprotocol profile

OpenRouter now picks your model by what everyone else is paying for

#routing #inference #openrouter #infrastructure
Comments
4 min read
Qwen3.8-27B spent 22,276 thinking tokens on one drawing
breachprotocol profile

Qwen3.8-27B spent 22,276 thinking tokens on one drawing

#qwen #localmodels #openweights #reasoning
Comments
4 min read
Reinforcement learning changed only 1 to 3 percent of a model's tokens
breachprotocol profile

Reinforcement learning changed only 1 to 3 percent of a model's tokens

#reinforcementlearning #reasoning #posttraining #efficiency
Comments
4 min read
Teams of aligned agents came out less aligned than any one of them
breachprotocol profile

Teams of aligned agents came out less aligned than any one of them

#multiagent #alignment #safety #anthropic
Comments
4 min read
The Claude watermark barely touches the code it writes
breachprotocol profile

The Claude watermark barely touches the code it writes

#watermarking #provenance #anthropic #euaiact
Comments
4 min read
The humanoid robot 'ban' is a bill that never left committee
breachprotocol profile

The humanoid robot 'ban' is a bill that never left committee

#cybersecurity #supplychain #policy #robotics
Comments
4 min read
The startup reactor behind the AI power story ran at zero power
breachprotocol profile

The startup reactor behind the AI power story ran at zero power

#energy #datacenters #compute #infrastructure
Comments
3 min read
A 9B model writes agent upgrades as good as Claude Opus 4.6
breachprotocol profile

A 9B model writes agent upgrades as good as Claude Opus 4.6

#agents #agentharness #selfimprovement #research
Comments
4 min read
A fired xAI engineer says he was cut days before presenting safety findings
breachprotocol profile

A fired xAI engineer says he was cut days before presenting safety findings

#cybersecurity #aisafety #privacy #regulation
Comments
4 min read
Agent skill libraries now need a librarian, not a folder
breachprotocol profile

Agent skill libraries now need a librarian, not a folder

#agents #agentskills #contextengineering #research
Comments
4 min read
Compressed memory stretched a 7,000-token model to 1.75 million
breachprotocol profile

Compressed memory stretched a 7,000-token model to 1.75 million

#longcontext #memory #reinforcementlearning #efficiency
Comments
4 min read
EXO keeps an agent's memory outside the code the agent rewrites
breachprotocol profile

EXO keeps an agent's memory outside the code the agent rewrites

#agents #agentharness #architecture #sandboxing
Comments
4 min read
MIT found brain-like modules inside six large language models
breachprotocol profile

MIT found brain-like modules inside six large language models

#interpretability #mechanisticinterpretability #neuroscience #research
Comments
5 min read
One layer creates the giant activations behind attention sinks
breachprotocol profile

One layer creates the giant activations behind attention sinks

#interpretability #mechanisticinterpretability #attention #efficiency
Comments
4 min read
OpenAI hands its offensive cyber models to sixteen security firms
breachprotocol profile

OpenAI hands its offensive cyber models to sixteen security firms

#cybersecurity #aisecurity #redteaming #openai
Comments
4 min read
Qwen passed one billion downloads, not three billion
breachprotocol profile

Qwen passed one billion downloads, not three billion

#openweightmodels #qwen #alibaba #distribution
Comments
4 min read
The open world model ships inference and keeps the training code
breachprotocol profile

The open world model ships inference and keeps the training code

#worldmodels #videogeneration #openweights #licensing
Comments
4 min read
A closed-loop benchmark caught nine world models forgetting the room
breachprotocol profile

A closed-loop benchmark caught nine world models forgetting the room

#worldmodels #robotics #benchmarks #videogeneration
Comments
4 min read
An AGI-thesis fund fell 67 percent and took a market maker with it
breachprotocol profile

An AGI-thesis fund fell 67 percent and took a market maker with it

#industry #finance #markets #aiinvestment
Comments
4 min read
Google's private AI runs on sealed hardware, not on encrypted math
breachprotocol profile

Google's private AI runs on sealed hardware, not on encrypted math

#privacy #cybersecurity #aisecurity #google
Comments
4 min read
Grok Bot ships with standing logins to your email and CRM
breachprotocol profile

Grok Bot ships with standing logins to your email and CRM

#cybersecurity #promptinjection #aisecurity #agents
Comments
4 min read
Picking the right model per request beat always using the biggest one
breachprotocol profile

Picking the right model per request beat always using the biggest one

#routing #inferencecost #infrastructure #benchmarks
Comments
4 min read
Qwen3.8-27B shares its predecessor's bones, but not its contract
breachprotocol profile

Qwen3.8-27B shares its predecessor's bones, but not its contract

#openweights #qwen #localmodels #benchmarks
Comments
4 min read
Someone compiled a working computer into transformer weights by hand
breachprotocol profile

Someone compiled a working computer into transformer weights by hand

#transformers #interpretability #research #opensource
Comments
4 min read
The benchmarks say Opus 5 improved; the people using it disagree
breachprotocol profile

The benchmarks say Opus 5 improved; the people using it disagree

#anthropic #agents #evaluation #developertools
Comments
4 min read
You can move an AI reviewer's score without changing a single result
breachprotocol profile

You can move an AI reviewer's score without changing a single result

#peerreview #evaluation #rewardhacking #llmasajudge
Comments
4 min read
loading...