One shared ChatGPT link took over the signed-in account
A public ChatGPT share ran attacker JS as the signed-in origin. Parameter's write-up, OpenAI's fix, and a local rebuild that shows which check stops it.
Tags
62 posts tagged
A public ChatGPT share ran attacker JS as the signed-in origin. Parameter's write-up, OpenAI's fix, and a local rebuild that shows which check stops it.
"As more people delegate writing to LLMs they will also delegate thinking and comprehension to LLMs. I don't want to do that, where's the fun of writing then?"
— Martin Fowler
Mistral Large 4's open weights are 1,060.6 GB in FP8 and 610 GB in NVFP4. What fits on one 8-GPU node, KV cache per token, and API vs renting GPUs.
A safetensors file starts with an 8-byte header length and a JSON header. Two Range requests per shard counted Mistral Large 4's 1,053B params from 6.2 MB.
Gemini 4 Carbon will cost $4/$20 per million tokens after its intro price, the same as Opus 5.5. Its implicit-only cache decides which is cheaper.
Surge AI's DAYJOB: Tax tests agents on 50 messy tax assignments. Sonnet 5.5 leads at 10.8%. Tasks reward spotting issues the client never raised.
DAYJOB: Tax is Surge AI's new benchmark for AI agents doing professional tax work, published October 27, 2026. It has 50 assignments across eight areas of tax p...
via Hacker NewsAnthropic gave 50 analysts' Claude Code agents a hidden coin-flip lean. Analysts matched it 86% of the time and 15% credited the agent.
How AI agents shape experts' judgment calls is a new study from Anthropic's Alignment team (October 27, 2026). They gave 50 professional analysts a realistic ca...
OpenAI released an AI-written disproof of the Exponential Time Hypothesis with a Lean proof, shepherded by the researchers who proposed ETH.
OpenAI's Working with academic communities to build scientific understanding (October 26, 2026) announces three results its internal models produced: a proof of...
Anthropic's S-1: $16.3B revenue in H1 2026, 80% of it metered tokens, and cost per task named as the lever it competes on. Notes from the filing.
Anthropic, PBC: Form S-1 registration statement (filed with the SEC on October 26, 2026) Anthropic filed to go public on the Nasdaq Global Select Market under t...
via Hacker NewsA KV cache pool smaller than the context window made Cloudflare's GLM-5.3 return other users' data in 155 responses. How it happened and the SGLang fix.
Cross-request KV cache exposure in the GLM-5.3 Workers AI model, and our remediation with SGLang (Kevin Flansburg and Michelle Chen, Cloudflare, October 26, 202...
via Hacker News"If you care about quality only, pick Astra. If you're a bit price sensitive, pick Sol. If you're very price sensitive, pick Terra or Luna."
— Ted Sanders
Apollo's chief economist sees no AI payoff outside big tech, and says companies now treat token spend as a cost center. What builders should do.
No Signs of AI Payoff Outside the Tech Sector (Torsten Slok, Apollo's chief economist, The Daily Spark, October 24, 2026) Slok's note is five charts and four sh...
via Hacker NewsA 2015 Inception v3 CNN beat GPT-6 Astra and Claude Fable 5.1 at diagnosing 200 skin lesions, 62% to 42%. What it says about small models.
A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis (Dr Magnus Lynch, October 25, 2026) Magnus Lynch, a consultant dermatologist at Guy's Hosp...
via Hacker NewsMax Woolf's timeline of OpenAI and Anthropic price cuts, and why cache-read prices are the ones that matter for coding agents.
OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why? (Max Woolf, October 25, 2026) Max Woolf lays out every OpenAI and Anthropic pri...
via Hacker News"It's a simple question for any enterprise. Would you like something that's 10 percent of the cost if it's 90 percent as good as the best, or do you want to pay that 10 times higher cost?"
— Arvind Krishna
Kimi K3.1 Pro lists at half Opus 5.5's price, but cache hits narrow the gap. Session cost math, Moonshot's benchmarks, and the cyber data it removed.
Block's Dessa is a full-duplex speech model trained only on human conversation audio. What it gains by skipping text, and what it can't do yet.
Dessa: the Conversation Model trained only on speech (via Hacker News) Dessa is a speech-to-speech model from a small research lab inside Block, announced on Oc...
via Hacker News (295 points on 2026-10-24)OpenAI's official CLI has a tokenizer command that counts and inspects tokens locally, with no API key or network. Commands and real output.
Tokens are the chunks of text an AI model reads and bills you for. See how tokenizers split text, why Gujarati costs more, and how to count them.
"When all is said and told, the "naturalness" with which we use our native tongues boils down to the ease with which we can use them for making statements the nonsense of which is not obvious."
— Edsger W. Dijkstra
LLM temperature controls how random the next token is. How temperature, top-p and top-k reshape the probabilities, and why Claude's API dropped them in 2026.
How to choose an LLM: set hard constraints, test candidates on your own eval set, then compare cost per solved task, not price per token. With a cost script.
The EU AI Act for developers: risk tiers, provider vs deployer, Article 50 transparency, GPAI rules and the AI Omnibus dates, with a Python triage script.
A system prompt is text placed before the conversation that sets a model's role and rules. See the raw chat template, its token cost and its limits.
Quantization stores model weights in fewer bits plus a scale. Runnable int8, int4, MXFP4 and NVFP4 code, the memory math, and where accuracy goes.
An AI agent is a model that calls tools in a loop until a goal is met. A runnable Python loop, what each step costs in tokens, and where agents break.
Sparse attention lets each token attend to a few chosen earlier tokens. How DeepSeek V3.2 does it, why long prompts get cheaper, and a runnable numpy demo.
Claude Opus 4.5 costs $5/$25 per million tokens and adds a beta effort setting. How effort changes token spend and cost per task, with a runnable script.
A context window is how many tokens a model handles per request. Gemini 3 Pro takes 1M, but past 200k each token costs double. Runnable cost script inside.
Reasoning models write hidden thinking tokens before answering, billed as output. How GPT-5.1's reasoning effort works, with a runnable cost script.
ChatGPT Atlas puts an agent inside your browser. How prompt injection hides in page text, why filtering it fails, and where the real check has to live.
LLMs hallucinate because training and benchmarks reward a confident guess over 'I don't know'. Where hallucinations come from and how to cut them, in Python.
Mechanistic interpretability reverse-engineers neural networks into features and circuits. Superposition, sparse autoencoders and a Python toy.
DeepSeek released R1 on 20 January 2025 under MIT, at $0.55 per million input tokens, with 79.8% on AIME 2024. What that does to a reasoning-search bill.
A history of large language models, from the 2017 transformer and GPT-3 to reasoning models and agents in 2026, with dates, sources and training numbers.
Open-weight models publish their trained weights but not always the data or code. How that differs from open source AI, and why it matters to builders.
Test-time compute is the work a model does while answering: longer reasoning, more samples, a verifier. How it scales and what it costs, with Python.
o1-preview launched 12 September 2024. Thinking is billed as output, the chain of thought stays hidden, and the IMO score in the post is not the preview.
Structured output forces an LLM's reply to match a JSON schema by masking invalid tokens during decoding. JSON mode vs strict mode, with a runnable Python demo.
Scaling laws predict how an LLM's loss falls as you add parameters, data and compute. What Kaplan and Chinchilla found, and why models now overtrain.
LLM routing sends each query to the cheapest model likely to answer it well; cascades try a small model first and escalate. How both work, with a Python sim.
Claude 3.5 Sonnet launched 21 June 2024 at $3 and $15 per million tokens, with Artifacts and a 64% internal coding score.
LLM benchmarks are fixed test sets that score models on knowledge, reasoning, coding or agents. What the main ones measure, why they saturate, how to read them.
Multimodal models read images, audio or video next to text. How vision language models turn pixels into tokens, what an image costs, with a Python demo.
GPT-4o launched 13 May 2024 at $5 and $15 per million tokens, half of GPT-4 Turbo, with text and vision in the API.
Small language models are LLMs of about 1B to 14B parameters that run on a phone or laptop. How they are made, how much memory they need, and when to use one.
Google announced Gemini 1.5 on 15 February 2024 with a 128k default and a 1 million token preview. What the needle test measures, and what it leaves out.
Mixture of experts (MoE) routes each token to a few expert networks, so a model can hold 47B parameters but compute with 13B. How routing works, in Python.
The KV cache stores each token's attention keys and values so an LLM doesn't recompute them per new token. How it works, how big it gets, and GQA.
RoPE (rotary position embeddings) encodes token position by rotating query and key vectors, so attention sees relative distance. How it works, in Python.
AI jailbreaking tricks a model into ignoring its safety training. Why it works, how it differs from prompt injection, and which defenses hold, in Python.
Meta released Llama 2 on 18 July 2023 in 7B, 13B, and 70B sizes for research and commercial use. The license lines to read before you download weights.
On 13 June 2023, GPT-4 and GPT-3.5 Turbo started returning JSON for functions you describe. The schema, the injection caveat, and an offline check.
RAG (retrieval-augmented generation) fetches relevant text from your data and puts it in the prompt so the LLM answers from it. How it works, in Python.
Prompt injection is when untrusted text in an LLM's context overrides your instructions. Direct vs indirect attacks, why filters fail, and designs that hold.
GPT-4 launched 14 March 2023 at $30 per million input tokens for an 8,192 context. What the system card admits, priced against the 32k window.
How LLMs are trained: pretraining on trillions of tokens, then supervised fine-tuning, RLHF or DPO, and RL with verifiable rewards. With a compute calculator.
Attention lets each token in a transformer weigh every other token. How query, key and value vectors work, why the sqrt(d_k) is there, with runnable Python.
Embeddings turn text into vectors of numbers so that similar meanings sit close together. How they are made and compared, with a runnable Python demo.
A transformer is the neural network architecture behind modern LLMs. How its layers, attention and feed-forward blocks work, with a Python parameter counter.
A large language model is a neural network trained to predict the next token of text. How LLMs work, what 'large' means, and a 25-line Python demo.
ChatGPT launched 30 November 2022 as a free research preview of a GPT-3.5 model tuned with RLHF. The limits OpenAI printed on day one.