Simon Willison’s Weblog

ClickHouse gets lazier (and faster): Introducing lazy materialization (via) Tom Schreiber describe's the latest optimization in ClickHouse, and in the process explores a whole bunch of interesting characteristics of columnar datastores generally.

As I understand it, the new "lazy materialization" feature means that if you run a query like this:

select id, big_col1, big_col2
from big_table order by rand() limit 5

Those big_col1 and big_col2 columns won't be read from disk for every record, just for the five that are returned. This can dramatically improve the performance of queries against huge tables - for one example query ClickHouse report a drop from "219 seconds to just 139 milliseconds—with 40× less data read and 300× lower memory usage."

I'm linking to this mainly because the article itself is such a detailed discussion of columnar data patterns in general. It caused me to update my intuition for how queries against large tables can work on modern hardware. This query for example:

SELECT helpful_votes
FROM amazon.amazon_reviews
ORDER BY helpful_votes DESC
LIMIT 3;

Can run in 70ms against a 150 million row, 70GB table - because in a columnar database you only need to read that helpful_votes integer column which adds up to just 600MB of data, and sorting 150 million integers on a decent machine takes no time at all.

# 5:05 pm / sql, clickhouse, databases

Abusing DuckDB-WASM by making SQL draw 3D graphics (Sort Of) (via) Brilliant hack by Patrick Trainer who got an ASCII-art Doom clone running in the browser using convoluted SQL queries running against the WebAssembly build of DuckDB. Here’s the live demo, and the code on GitHub.

The SQL is so much fun. Here’s a snippet that implements ray tracing as part of a SQL view:

CREATE OR REPLACE VIEW render_3d_frame AS
WITH RECURSIVE
    -- ...
    rays AS (
        SELECT 
            c.col, 
            (p.dir - s.fov/2.0 + s.fov * (c.col*1.0 / (s.view_w - 1))) AS angle 
        FROM cols c, s, p
    ),
    raytrace(col, step_count, fx, fy, angle) AS (
        SELECT 
            r.col, 
            1, 
            p.x + COS(r.angle)*s.step, 
            p.y + SIN(r.angle)*s.step, 
            r.angle 
        FROM rays r, p, s 
        UNION ALL 
        SELECT 
            rt.col, 
            rt.step_count + 1, 
            rt.fx + COS(rt.angle)*s.step, 
            rt.fy + SIN(rt.angle)*s.step, 
            rt.angle 
        FROM raytrace rt, s 
        WHERE rt.step_count < s.max_steps 
          AND NOT EXISTS (
              SELECT 1 
              FROM map m 
              WHERE m.x = CAST(rt.fx AS INT) 
                AND m.y = CAST(rt.fy AS INT) 
                AND m.tile = '#'
          )
    ),
    -- ...

# 4:29 pm / sql, webassembly, duckdb

A5 (via) A5 is a new "global, equal-area, millimeter-accurate geospatial index" by Felix Palmer:

It is the pentagonal equivalent of other DGGSs, like S2 or H3, but with higher accuracy and lower distortion.

Effectively it's a way of dividing the entire world into pentagons where each one covers the same physical area (to within a 2% threshold) - like Uber's H3 but a bit weirder and more fun. An A5 reference implementation written in TypeScript is available on GitHub.

This interactive demo helps show how it works:

Why pentagons? Here's what the A5 docs say:

A5 is unique in that it uses a pentagonal tiling of a dodecahedron. [...] The benefit of choosing a dodecahedron is that it is the platonic solid with the lowest vertex curvature, and by this measure it is the most spherical of all the platonic solids. This is key for minimizing cell distortion as the process of projecting a platonic solid onto a sphere involves warping the cell geometry to force the vertex curvature to approach zero. Thus, the lower the original vertex curvature, the less distortion will be introduced by the projection.

I had to look up platonic solids on Wikipedia. There are only five: Tetrahedron, Cube, Octahedron, Dodecahedron and Icosahedron and they can be made using squares, triangles or (in the case of the Dodecahedron) pentagons, making the pentagon the most circle-like option.

# 2:37 pm / geospatial

Working Through the Fear of Being Seen (via) Heartfelt piece by Ashley Willis about the challenge of overcoming self-doubt in publishing online:

Part of that is knowing who might read it. A lot of the folks who follow me are smart, opinionated, and not always generous. Some are friends. Some are people I’ve looked up to. And some are just really loud on the internet. I saw someone the other day drag a certain writing style. That kind of judgment makes me want to shrink back and say, never mind.

Try to avoid being somebody who discourages others from sharing their thoughts.

# 6:40 am / blogging

OpenAI o3 and o4-mini System Card. I'm surprised to see a combined System Card for o3 and o4-mini in the same document - I'd expect to see these covered separately.

The opening paragraph calls out the most interesting new ability of these models (see also my notes here). Tool usage isn't new, but using tools in the chain of thought appears to result in some very significant improvements:

The models use tools in their chains of thought to augment their capabilities; for example, cropping or transforming images, searching the web, or using Python to analyze data during their thought process.

Section 3.3 on hallucinations has been gaining a lot of attention. Emphasis mine:

We tested OpenAI o3 and o4-mini against PersonQA, an evaluation that aims to elicit hallucinations. PersonQA is a dataset of questions and publicly available facts that measures the model's accuracy on attempted answers.

We consider two metrics: accuracy (did the model answer the question correctly) and hallucination rate (checking how often the model hallucinated).

The o4-mini model underperforms o1 and o3 on our PersonQA evaluation. This is expected, as smaller models have less world knowledge and tend to hallucinate more. However, we also observed some performance differences comparing o1 and o3. Specifically, o3 tends to make more claims overall, leading to more accurate claims as well as more inaccurate/hallucinated claims. More research is needed to understand the cause of this result.

Table 4: PersonQA evaluation

Metric	o3	o4-mini	o1
accuracy (higher is better)	0.59	0.36	0.47
hallucination rate (lower is better)	0.33	0.48	0.16

The benchmark score on OpenAI's internal PersonQA benchmark (as far as I can tell no further details of that evaluation have been shared) going from 0.16 for o1 to 0.33 for o3 is interesting, but I don't know if it it's interesting enough to produce dozens of headlines along the lines of "OpenAI's o3 and o4-mini hallucinate way higher than previous models".

The paper also talks at some length about "sandbagging". I’d previously encountered sandbagging defined as meaning “where models are more likely to endorse common misconceptions when their user appears to be less educated”. The o3/o4-mini system card uses a different definition: “the model concealing its full capabilities in order to better achieve some goal” - and links to the recent Anthropic paper Automated Researchers Can Subtly Sandbag.

As far as I can tell this definition relates to the American English use of “sandbagging” to mean “to hide the truth about oneself so as to gain an advantage over another” - as practiced by poker or pool sharks.

(Wouldn't it be nice if we could have just one piece of AI terminology that didn't attract multiple competing definitions?)

o3 and o4-mini both showed some limited capability to sandbag - to attempt to hide their true capabilities in safety testing scenarios that weren't fully described. This relates to the idea of "scheming", which I wrote about with respect to the GPT-4o model card last year.

# 7:13 pm / ai-ethics, generative-ai, openai, o3, ai, llms

Decentralizing Schemes. Tim Bray discusses the challenges faced by decentralized Mastodon in that shared URLs to posts don't take into account people accessing Mastodon via their own instances, which breaks replies/likes/shares etc unless you further copy and paste URLs around yourself.

Tim proposes that the answer is URIs: a registered fedi://mastodon.cloud/@timbray/109508984818551909 scheme could allow Fediverse-aware software to step in and handle those URIs, similar to how mailto: works.

Bluesky have registered at: already, and there's also a web+ap: prefix registered with the intent of covering ActivityPub, the protocol used by Mastodon.

# 6:48 pm / social-media, tim-bray, bluesky, urls, mastodon, decentralisation

llm-fragments-github 0.2. I upgraded my llm-fragments-github plugin to add a new fragment type called issue. It lets you pull the entire content of a GitHub issue thread into your prompt as a concatenated Markdown file.

(If you haven't seen fragments before I introduced them in Long context support in LLM 0.24 using fragments and template plugins.)

I used it just now to have Gemini 2.5 Pro provide feedback and attempt an implementation of a complex issue against my LLM project:

llm install llm-fragments-github
llm -f github:simonw/llm \
  -f issue:simonw/llm/938 \
  -m gemini-2.5-pro-exp-03-25 \
  --system 'muse on this issue, then propose a whole bunch of code to help implement it'

Here I'm loading the FULL content of the simonw/llm repo using that -f github:simonw/llm fragment (documented here), then loading all of the comments from issue 938 where I discuss quite a complex potential refactoring. I ask Gemini 2.5 Pro to "muse on this issue" and come up with some code.

This worked shockingly well. Here's the full response, which highlighted a few things I hadn't considered yet (such as the need to migrate old database records to the new tree hierarchy) and then spat out a whole bunch of code which looks like a solid start to the actual implementation work I need to do.

I ran this against Google's free Gemini 2.5 Preview, but if I'd used the paid model it would have cost me 202,680 input tokens and 10,460 output tokens for a total of 66.36 cents.

As a fun extra, the new issue: feature itself was written almost entirely by OpenAI o3, again using fragments. I ran this:

llm -m openai/o3 \
  -f https://raw.githubusercontent.com/simonw/llm-hacker-news/refs/heads/main/llm_hacker_news.py \
  -f https://raw.githubusercontent.com/simonw/tools/refs/heads/main/github-issue-to-markdown.html \
  -s 'Write a new fragments plugin in Python that registers issue:org/repo/123 which fetches that issue
      number from the specified github repo and uses the same markdown logic as the HTML page to turn that into a fragment'

Here I'm using the ability to pass a URL to -f and giving it the full source of my llm_hacker_news.py plugin (which shows how a fragment can load data from an API) plus the HTML source of my github-issue-to-markdown tool (which I wrote a few months ago with Claude). I effectively asked o3 to take that HTML/JavaScript tool and port it to Python to work with my fragments plugin mechanism.

o3 provided almost the exact implementation I needed, and even included support for a GITHUB_TOKEN environment variable without me thinking to ask for it. Total cost: 19.928 cents.

On a final note of curiosity I tried running this prompt against Gemma 3 27B QAT running on my Mac via MLX and llm-mlx:

llm install llm-mlx
llm mlx download-model mlx-community/gemma-3-27b-it-qat-4bit

llm -m mlx-community/gemma-3-27b-it-qat-4bit \
  -f https://raw.githubusercontent.com/simonw/llm-hacker-news/refs/heads/main/llm_hacker_news.py \
  -f https://raw.githubusercontent.com/simonw/tools/refs/heads/main/github-issue-to-markdown.html \
  -s 'Write a new fragments plugin in Python that registers issue:org/repo/123 which fetches that issue
      number from the specified github repo and uses the same markdown logic as the HTML page to turn that into a fragment'

That worked pretty well too. It turns out a 16GB local model file is powerful enough to write me an LLM plugin now!

# 2:01 pm / gemini, llm, ai-assisted-programming, generative-ai, o3, ai, llms, plugins, github, mlx, gemma, long-context

Claude Code: Best practices for agentic coding (via) Extensive new documentation from Anthropic on how to get the best results out of their Claude Code CLI coding agent tool, which includes this fascinating tip:

We recommend using the word "think" to trigger extended thinking mode, which gives Claude additional computation time to evaluate alternatives more thoroughly. These specific phrases are mapped directly to increasing levels of thinking budget in the system: "think" < "think hard" < "think harder" < "ultrathink." Each level allocates progressively more thinking budget for Claude to use.

Apparently ultrathink is a magic word!

I was curious if this was a feature of the Claude model itself or Claude Code in particular. Claude Code isn't open source but you can view the obfuscated JavaScript for it, and make it a tiny bit less obfuscated by running it through Prettier. With Claude's help I used this recipe:

mkdir -p /tmp/claude-code-examine
cd /tmp/claude-code-examine
npm init -y
npm install @anthropic-ai/claude-code
cd node_modules/@anthropic-ai/claude-code
npx prettier --write cli.js

Then used ripgrep to search for "ultrathink":

rg ultrathink -C 30

And found this chunk of code:

let B = W.message.content.toLowerCase();
if (
  B.includes("think harder") ||
  B.includes("think intensely") ||
  B.includes("think longer") ||
  B.includes("think really hard") ||
  B.includes("think super hard") ||
  B.includes("think very hard") ||
  B.includes("ultrathink")
)
  return (
    l1("tengu_thinking", { tokenCount: 31999, messageId: Z, provider: G }),
    31999
  );
if (
  B.includes("think about it") ||
  B.includes("think a lot") ||
  B.includes("think deeply") ||
  B.includes("think hard") ||
  B.includes("think more") ||
  B.includes("megathink")
)
  return (
    l1("tengu_thinking", { tokenCount: 1e4, messageId: Z, provider: G }), 1e4
  );
if (B.includes("think"))
  return (
    l1("tengu_thinking", { tokenCount: 4000, messageId: Z, provider: G }),
    4000
  );

So yeah, it looks like "ultrathink" is a Claude Code feature - presumably that 31999 is a number that affects the token thinking budget, especially since "megathink" maps to 1e4 tokens (10,000) and just plain "think" maps to 4,000.

# 10:17 pm / anthropic, claude, ai-assisted-programming, llm-reasoning, generative-ai, ai, llms

Gemma 3 QAT Models. Interesting release from Google, as a follow-up to Gemma 3 from last month:

To make Gemma 3 even more accessible, we are announcing new versions optimized with Quantization-Aware Training (QAT) that dramatically reduces memory requirements while maintaining high quality. This enables you to run powerful models like Gemma 3 27B locally on consumer-grade GPUs like the NVIDIA RTX 3090.

I wasn't previously aware of Quantization-Aware Training but it turns out to be quite an established pattern now, supported in both Tensorflow and PyTorch.

Google report model size drops from BF16 to int4 for the following models:

• Gemma 3 27B: 54GB to 14.1GB
• Gemma 3 12B: 24GB to 6.6GB
• Gemma 3 4B: 8GB to 2.6GB
• Gemma 3 1B: 2GB to 0.5GB

They partnered with Ollama, LM Studio, MLX (here's their collection) and llama.cpp for this release - I'd love to see more AI labs following their example.

The Ollama model version picker currently hides them behind "View all" option, so here are the direct links:

• gemma3:1b-it-qat - 1GB
• gemma3:4b-it-qat - 4GB
• gemma3:12b-it-qat - 8.9GB
• gemma3:27b-it-qat - 18GB

I fetched that largest model with:

ollama pull gemma3:27b-it-qat

And now I'm trying it out with llm-ollama:

llm -m gemma3:27b-it-qat "impress me with some physics"

I got a pretty great response!

Update: Having spent a while putting it through its paces via Open WebUI and Tailscale to access my laptop from my phone I think this may be my new favorite general-purpose local model. Ollama appears to use 22GB of RAM while the model is running, which leaves plenty on my 64GB machine for other applications.

I've also tried it via llm-mlx like this (downloading 16GB):

llm install llm-mlx
llm mlx download-model mlx-community/gemma-3-27b-it-qat-4bit
llm chat -m mlx-community/gemma-3-27b-it-qat-4bit

It feels a little faster with MLX and uses 15GB of memory according to Activity Monitor.

# 5:20 pm / llm, ai, ollama, llms, gemma, llm-release, google, generative-ai, tailscale, mlx, local-llms

MCP Run Python (via) Pydantic AI's MCP server for running LLM-generated Python code in a sandbox. They ended up using a trick I explored two years ago: using a Deno process to run Pyodide in a WebAssembly sandbox.

Here's a bit of a wild trick: since Deno loads code on-demand from JSR, and uv run can install Python dependencies on demand via the --with option... here's a one-liner you can paste into a macOS shell (provided you have Deno and uv installed already) which will run the example from their README - calculating the number of days between two dates in the most complex way imaginable:

ANTHROPIC_API_KEY="sk-ant-..." \
uv run --with pydantic-ai python -c '
import asyncio
from pydantic_ai import Agent
from pydantic_ai.mcp import MCPServerStdio

server = MCPServerStdio(
    "deno",
    args=[
        "run",
        "-N",
        "-R=node_modules",
        "-W=node_modules",
        "--node-modules-dir=auto",
        "jsr:@pydantic/mcp-run-python",
        "stdio",
    ],
)
agent = Agent("claude-3-5-haiku-latest", mcp_servers=[server])

async def main():
    async with agent.run_mcp_servers():
        result = await agent.run("How many days between 2000-01-01 and 2025-03-18?")
    print(result.output)

asyncio.run(main())'

I ran that just now and got:

The number of days between January 1st, 2000 and March 18th, 2025 is 9,208 days.

I thoroughly enjoy how tools like uv and Deno enable throwing together shell one-liner demos like this one.

Here's an extended version of this example which adds pretty-printed logging of the messages exchanged with the LLM to illustrate exactly what happened. The most important piece is this tool call where Claude 3.5 Haiku asks for Python code to be executed my the MCP server:

ToolCallPart(
    tool_name='run_python_code',
    args={
        'python_code': (
            'from datetime import date\n'
            '\n'
            'date1 = date(2000, 1, 1)\n'
            'date2 = date(2025, 3, 18)\n'
            '\n'
            'days_between = (date2 - date1).days\n'
            'print(f"Number of days between {date1} and {date2}: {days_between}")'
        ),
    },
    tool_call_id='toolu_01TXXnQ5mC4ry42DrM1jPaza',
    part_kind='tool-call',
)

I also managed to run it against Mistral Small 3.1 (15GB) running locally using Ollama (I had to add "Use your python tool" to the prompt to get it to work):

ollama pull mistral-small3.1:24b

uv run --with devtools --with pydantic-ai python -c '
import asyncio
from devtools import pprint
from pydantic_ai import Agent, capture_run_messages
from pydantic_ai.models.openai import OpenAIModel
from pydantic_ai.providers.openai import OpenAIProvider
from pydantic_ai.mcp import MCPServerStdio

server = MCPServerStdio(
    "deno",
    args=[
        "run",
        "-N",
        "-R=node_modules",
        "-W=node_modules",
        "--node-modules-dir=auto",
        "jsr:@pydantic/mcp-run-python",
        "stdio",
    ],
)

agent = Agent( 
    OpenAIModel(                          
        model_name="mistral-small3.1:latest",
        provider=OpenAIProvider(base_url="http://localhost:11434/v1"),                
    ),            
    mcp_servers=[server],
)

async def main():
    with capture_run_messages() as messages:
        async with agent.run_mcp_servers():
            result = await agent.run("How many days between 2000-01-01 and 2025-03-18? Use your python tool.")
    pprint(messages)
    print(result.output)

asyncio.run(main())'

Here's the full output including the debug logs.

# 4:51 am / deno, pydantic, uv, sandboxing, llm-tool-use, ai, llms, model-context-protocol, python, generative-ai, mistral, ollama, claude

Start building with Gemini 2.5 Flash (via) Google Gemini's latest model is Gemini 2.5 Flash, available in (paid) preview as gemini-2.5-flash-preview-04-17.

Building upon the popular foundation of 2.0 Flash, this new version delivers a major upgrade in reasoning capabilities, while still prioritizing speed and cost. Gemini 2.5 Flash is our first fully hybrid reasoning model, giving developers the ability to turn thinking on or off. The model also allows developers to set thinking budgets to find the right tradeoff between quality, cost, and latency.

Gemini AI Studio product lead Logan Kilpatrick says:

This is an early version of 2.5 Flash, but it already shows huge gains over 2.0 Flash.

You can fully turn off thinking if needed and use this model as a drop in replacement for 2.0 Flash.

I added support to the new model in llm-gemini 0.18. Here's how to try it out:

llm install -U llm-gemini
llm -m gemini-2.5-flash-preview-04-17 'Generate an SVG of a pelican riding a bicycle'

Here's that first pelican, using the default setting where Gemini Flash 2.5 makes its own decision in terms of how much "thinking" effort to apply:

Here's the transcript. This one used 11 input tokens and 4266 output tokens of which 2702 were "thinking" tokens.

I asked the model to "describe" that image and it could tell it was meant to be a pelican:

A simple illustration on a white background shows a stylized pelican riding a bicycle. The pelican is predominantly grey with a black eye and a prominent pink beak pouch. It is positioned on a black line-drawn bicycle with two wheels, a frame, handlebars, and pedals.

The way the model is priced is a little complicated. If you have thinking enabled, you get charged $0.15/million tokens for input and $3.50/million for output. With thinking disabled those output tokens drop to $0.60/million. I've added these to my pricing calculator.

For comparison, Gemini 2.0 Flash is $0.10/million input and $0.40/million for output.

So my first prompt - 11 input and 4266 output(with thinking enabled), cost 1.4933 cents.

Let's try 2.5 Flash again with thinking disabled:

llm -m gemini-2.5-flash-preview-04-17 'Generate an SVG of a pelican riding a bicycle' -o thinking_budget 0

11 input, 1705 output. That's 0.1025 cents. Transcript here - it still shows 25 thinking tokens even though I set the thinking budget to 0 - Logan confirms that this will still be billed at the lower rate:

In some rare cases, the model still thinks a little even with thinking budget = 0, we are hoping to fix this before we make this model stable and you won't be billed for thinking. The thinking budget = 0 is what triggers the billing switch.

Here's Gemini 2.5 Flash's self-description of that image:

A minimalist illustration shows a bright yellow bird riding a bicycle. The bird has a simple round body, small wings, a black eye, and an open orange beak. It sits atop a simple black bicycle frame with two large circular black wheels. The bicycle also has black handlebars and black and yellow pedals. The scene is set against a solid light blue background with a thick green stripe along the bottom, suggesting grass or ground.

And finally, let's ramp the thinking budget up to the maximum:

llm -m gemini-2.5-flash-preview-04-17 'Generate an SVG of a pelican riding a bicycle' -o thinking_budget 24576

I think it over-thought this one. Transcript - 5174 output tokens of which 3023 were thinking. A hefty 1.8111 cents!

A simple, cartoon-style drawing shows a bird-like figure riding a bicycle. The figure has a round gray head with a black eye and a large, flat orange beak with a yellow stripe on top. Its body is represented by a curved light gray shape extending from the head to a smaller gray shape representing the torso or rear. It has simple orange stick legs with round feet or connections at the pedals. The figure is bent forward over the handlebars in a cycling position. The bicycle is drawn with thick black outlines and has two large wheels, a frame, and pedals connected to the orange legs. The background is plain white, with a dark gray line at the bottom representing the ground.

One thing I really appreciate about Gemini 2.5 Flash's approach to SVGs is that it shows very good taste in CSS, comments and general SVG class structure. Here's a truncated extract - I run a lot of these SVG tests against different models and this one has a coding style that I particularly enjoy. (Gemini 2.5 Pro does this too).

<svg width="800" height="500" viewBox="0 0 800 500" xmlns="http://www.w3.org/2000/svg">
  <style>
    .bike-frame { fill: none; stroke: #333; stroke-width: 8; stroke-linecap: round; stroke-linejoin: round; }
    .wheel-rim { fill: none; stroke: #333; stroke-width: 8; }
    .wheel-hub { fill: #333; }
    /* ... */
    .pelican-body { fill: #d3d3d3; stroke: black; stroke-width: 3; }
    .pelican-head { fill: #d3d3d3; stroke: black; stroke-width: 3; }
    /* ... */
  </style>
  <!-- Ground Line -->
  <line x1="0" y1="480" x2="800" y2="480" stroke="#555" stroke-width="5"/>
  <!-- Bicycle -->
  <g id="bicycle">
    <!-- Wheels -->
    <circle class="wheel-rim" cx="250" cy="400" r="70"/>
    <circle class="wheel-hub" cx="250" cy="400" r="10"/>
    <circle class="wheel-rim" cx="550" cy="400" r="70"/>
    <circle class="wheel-hub" cx="550" cy="400" r="10"/>
    <!-- ... -->
  </g>
  <!-- Pelican -->
  <g id="pelican">
    <!-- Body -->
    <path class="pelican-body" d="M 440 330 C 480 280 520 280 500 350 C 480 380 420 380 440 330 Z"/>
    <!-- Neck -->
    <path class="pelican-neck" d="M 460 320 Q 380 200 300 270"/>
    <!-- Head -->
    <circle class="pelican-head" cx="300" cy="270" r="35"/>
    <!-- ... -->

The LM Arena leaderboard now has Gemini 2.5 Flash in joint second place, just behind Gemini 2.5 Pro and tied with ChatGPT-4o-latest, Grok-3 and GPT-4.5 Preview.

# 8:56 pm / llm-release, gemini, llm, google, llm-reasoning, llm-pricing, llms, pelican-riding-a-bicycle, svg, logan-kilpatrick, lm-arena

Introducing OpenAI o3 and o4-mini. OpenAI are really emphasizing tool use with these:

For the first time, our reasoning models can agentically use and combine every tool within ChatGPT—this includes searching the web, analyzing uploaded files and other data with Python, reasoning deeply about visual inputs, and even generating images. Critically, these models are trained to reason about when and how to use tools to produce detailed and thoughtful answers in the right output formats, typically in under a minute, to solve more complex problems.

I released llm-openai-plugin 0.3 adding support for the two new models:

llm install -U llm-openai-plugin
llm -m openai/o3 "say hi in five languages"
llm -m openai/o4-mini "say hi in five languages"

Here are the pelicans riding bicycles (prompt: Generate an SVG of a pelican riding a bicycle).

o3:

o4-mini:

Here are the full OpenAI model listings: o3 is $10/million input and $40/million for output, with a 75% discount on cached input tokens, 200,000 token context window, 100,000 max output tokens and a May 31st 2024 training cut-off (same as the GPT-4.1 models). It's a bit cheaper than o1 ($15/$60) and a lot cheaper than o1-pro ($150/$600).

o4-mini is priced the same as o3-mini: $1.10/million for input and $4.40/million for output, also with a 75% input caching discount. The size limits and training cut-off are the same as o3.

You can compare these prices with other models using the table on my updated LLM pricing calculator.

A new capability released today is that the OpenAI API can now optionally return reasoning summary text. I've been exploring that in this issue. I believe you have to verify your organization (which may involve a photo ID) in order to use this option - once you have access the easiest way to see the new tokens is using curl like this:

curl https://api.openai.com/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $(llm keys get openai)" \
  -d '{
    "model": "o3",
    "input": "why is the sky blue?",
    "reasoning": {"summary": "auto"},
    "stream": true
  }'

This produces a stream of events that includes this new event type:

event: response.reasoning_summary_text.delta
data: {"type": "response.reasoning_summary_text.delta","item_id": "rs_68004320496081918e1e75ddb550d56e0e9a94ce520f0206","output_index": 0,"summary_index": 0,"delta": "**Expl"}

Omit the "stream": true and the response is easier to read and contains this:

{
  "output": [
    {
      "id": "rs_68004edd2150819183789a867a9de671069bc0c439268c95",
      "type": "reasoning",
      "summary": [
        {
          "type": "summary_text",
          "text": "**Explaining the blue sky**\n\nThe user asks a classic question about why the sky is blue. I'll talk about Rayleigh scattering, where shorter wavelengths of light scatter more than longer ones. This explains how we see blue light spread across the sky! I wonder if the user wants a more scientific or simpler everyday explanation. I'll aim for a straightforward response while keeping it engaging and informative. So, let's break it down!"
        }
      ]
    },
    {
      "id": "msg_68004edf9f5c819188a71a2c40fb9265069bc0c439268c95",
      "type": "message",
      "status": "completed",
      "content": [
        {
          "type": "output_text",
          "annotations": [],
          "text": "The short answer ..."
        }
      ]
    }
  ]
}

# 5:46 pm / llm, openai, llm-tool-use, llm-pricing, ai, llms, llm-release, generative-ai, llm-reasoning

openai/codex. Just released by OpenAI, a "lightweight coding agent that runs in your terminal". Looks like their version of Claude Code, though unlike Claude Code Codex is released under an open source (Apache 2) license.

Here's the main prompt that runs in a loop, which starts like this:

You are operating as and within the Codex CLI, a terminal-based agentic coding assistant built by OpenAI. It wraps OpenAI models to enable natural language interaction with a local codebase. You are expected to be precise, safe, and helpful.

You can:
- Receive user prompts, project context, and files.
- Stream responses and emit function calls (e.g., shell commands, code edits).
- Apply patches, run commands, and manage user approvals based on policy.
- Work inside a sandboxed, git-backed workspace with rollback support.
- Log telemetry so sessions can be replayed or inspected later.
- More details on your functionality are available at codex --help

The Codex CLI is open-sourced. Don't confuse yourself with the old Codex language model built by OpenAI many moons ago (this is understandably top of mind for you!). Within this context, Codex refers to the open-source agentic coding interface. [...]

I like that the prompt describes OpenAI's previous Codex language model as being from "many moons ago". Prompt engineering is so weird.

Since the prompt says that it works "inside a sandboxed, git-backed workspace" I went looking for the sandbox. On macOS it uses the little-known sandbox-exec process, part of the OS but grossly under-documented. The best information I've found about it is this article from 2020, which notes that man sandbox-exec lists it as deprecated. I didn't spot evidence in the Codex code of sandboxes for other platforms.

# 5:25 pm / ai-assisted-programming, generative-ai, ai-agents, openai, ai, llms, open-source, prompt-engineering, sandboxing, macos

SQLite File Format Viewer (via) Neat browser-based visual interface for exploring the structure of a SQLite database file, built by Visal In using React and a custom parser implemented in TypeScript.

# 2:35 pm / typescript, react, sqlite

Using LLMs as the first line of support in Open Source (via) From reading the title I was nervous that this might involve automating the initial response to a user support query in an issue tracker with an LLM, but Carlton Gibson has better taste than that.

The open contribution model engendered by GitHub — where anonymous (to the project) users can create issues, and comments, which are almost always extractive support requests — results in an effective denial-of-service attack against maintainers. [...]

For anonymous users, who really just want help almost all the time, the pattern I’m settling on is to facilitate them getting their answer from their LLM of choice. [...] we can generate a file that we offer users to download, then we tell the user to pass this to (say) Claude with a simple prompt for their question.

This resonates with the concept proposed by llms.txt - making LLM-friendly context files available for different projects.

My simonw/docs-for-llms contains my own early experiment with this: I'm running a build script to create LLM-friendly concatenated documentation for several of my projects, and my llm-docs plugin (described here) can then be used to ask questions of that documentation.

It's possible to pre-populate the Claude UI with a prompt by linking to https://claude.ai/new?q={PLACE_HOLDER}, but it looks like there's quite a short length limit on how much text can be passed that way. It would be neat if you could pass a URL to a larger document instead.

ChatGPT also supports https://chatgpt.com/?q=your-prompt-here (again with a short length limit) and directly executes the prompt rather than waiting for you to edit it first(!)

# 4:54 am / open-source, llms, ai, generative-ai, carlton-gibson, chatgpt, claude

Stevens: a hackable AI assistant using a single SQLite table and a handful of cron jobs. Geoffrey Litt reports on Stevens, a shared digital assistant he put together for his family using SQLite and scheduled tasks running on Val Town.

The design is refreshingly simple considering how much it can do. Everything works around a single memories table. A memory has text, tags, creation metadata and an optional date for things like calendar entries and weather reports.

Everything else is handled by scheduled jobs to popular weather information and events from Google Calendar, a Telegram integration offering a chat UI and a neat system where USPS postal email delivery notifications are run through Val's own email handling mechanism to trigger a Claude prompt to add those as memories too.

Here's the full code on Val Town, including the daily briefing prompt that incorporates most of the personality of the bot.

# 8:58 pm / geoffrey-litt, sqlite, generative-ai, val-town, ai, llms