I think 'actual parallelism' is a vastly easier and more fruitful way to get better performance out of these kinds of systems, compared to pushing against single-threaded faster generation. Tool calling and responses are often embarrassingly parallel. Code generation tasks naturally have a dependency tree that can be unrolled into a fixed budget of parallelism. Tasks can be hierarchically decomposed into subtasks.
It's the same asynchronous stream pattern we're used to dealing with in regular software engineering. We have a fixed thread pool, lots of work that can be scheduled concurrently. Since these are streams, we can do the compute incrementally to reduce the time-to-first-byte/token/response.
Since so many tool calls are inherently asynchronous, and subagent task decomposition can be modelled as such, the IO streams can be oversubscribed, and incoming responses can be priority queued.
On the intelligence front, it's incredible how much better frontier models perform when you just interrupt them every so often and go 'is that the best you can do?', or re-iterate instructions, or repeat the overall goal. I find instruction following _so poor_, especially for 'presentation layer' aspects. Yet if I ask the model to rewrite its last response, it does so perfectly. Why can't the model do this 'internally' and save me having to say 'try again'!
Just because the 'model' is autoregressive doesn't mean the system as a whole needs to present a single stream of immutable text.
I don't buy the "any constraints cause lower performance via being out of distribution" idea. Sure if you ask the model to output 'reasoning' in JSON steps, that is a completely different 'channel' to its trained 'reasoning' output. For real tasks though, I think it's more about picking the _right_ context free grammar to enforce format correctness. You can enforce an in-distribution format and get the best of both worlds. I don't think the industry should settle so hard on JSON-for-everything.
I was working on a speculative decoding optimisation and its accompanying blog post. Explaining the more basic concepts filled so much of the post I decided to pull them out, forming this article.
I had a bit too much fun with the tokenisation diagrams / animations. The raw text is provided to an Astro component, which tokenises it, and forms the individual DOM elements of the tokens. I find it really hard to read 'tokenised' text, I figured some consistent colouring would help. The 'Probabilities' component is a trivial grid, but all the other components support 'word wrap'.
I ended up writing a 'responsive design aware graph colouring solver'.
Multiple screen widths, 'desktop' and 'mobile' are 'simulated', forming an adjacency graph of tokens that touch. Colours are then greedily allocated, then optimised per page over a few hundred iterations, swapping allocations to enforce minimum hue distance between touching tokens at those common screen sizes. The optimising value function prioritises even distribution of colours, because it looks nicer than maximal hue difference.
Originally I naively outputted the palette styles per component, but found the css post processing optimisers didn't handle that as well as I'd have thought. So then I wrote a little 'CSS compiler' that takes the high level palette and timing concepts of the animations, and optimally merges rule declarations.
The start of the post really relies on the animation occurring while fully in view, so I set up some IntersectionObservers that do the 'please scroll' text.
I tried my best to have it all work when JS is disabled on the client. I tried to get the 'hovering' to be CSS-only, but found the JS solution much more performant.
The DAG diagrams are formed with this neat Needleman-Wunsch algorithm from the bioinformatics field. The Astro component accepts several 'examples' then aligns common subsequences, producing the CSS grid and the 'basic SVG' on the server. The responsive nature meant I had to move the final 'allow' generation to the client.
Some browsers seem to throttle the token animations sometimes but I haven't figured out what causes that. This is my first time leaning hard on CSS variables.
In South Australia an algal bloom started in ~mid-March of this year, it's a pretty big ecological disaster, probably the worst non-bushfire disaster in living memory. Probably 30% of SA's coastline is affected. It's a pretty big deal affecting many people's livelihoods.
The joint state and federal government relief and cleanup package is worth AUD $102.5 million dollars.
I hope the public receives that comparison at every opportunity.
The old website was frankly excellent, the only problem was it didn't have HTTPS support. I would have happily upgraded that part of the system for the cost of a cup of coffee if I'd had an opportunity to submit for the tender!
The new website is significantly more difficult to navigate (for me, a seasoned tech user). The primary thing Dad's everywhere use it for (the weather radar) now requires scrolling to the _bottom_ of the page, and zooming in from the 'map of Australia' to the region you live in. It used to be like, a click to go from home page -> state weather radar with all the info you needed.
> [BOM] said the cost breakdown included $4.1 million for the redesign, $79.8 million for the website build, and the site's launch and security testing cost $12.6 million.
Absolutely stupid, even those numbers are outrageous. They say it's part of some 'larger upgrade package', prompted by a cyber attack in 2015.
But politicians over here love to blame cyber attacks when technical blunders happen. We had a census a couple years ago and the website fell over due to 'unprecedented load' or maybe it was a 'DDOS attack'? The news at the time couldn't decide who to blame!
Welp, I hope this gets as much world-wide attention as possible so they can be embarrassed and do better.
It would be nice if users of the codex-cli that are just using API keys as a way to handle rate limits and billing could receive these new models at the same time. I appreciate the reasoning behind delayed 'actual API' release, but I've found the rate limiting to be quite annoying, and my own API keys don't have this limitation.
Apart from the actual model release, this is the second set of models from OpenAI that uses the Harmony response format. I don't suppose anyone knows if OpenAI uses the Harmony format internally for GPT-5 as well?
It's the same asynchronous stream pattern we're used to dealing with in regular software engineering. We have a fixed thread pool, lots of work that can be scheduled concurrently. Since these are streams, we can do the compute incrementally to reduce the time-to-first-byte/token/response.
Since so many tool calls are inherently asynchronous, and subagent task decomposition can be modelled as such, the IO streams can be oversubscribed, and incoming responses can be priority queued.
On the intelligence front, it's incredible how much better frontier models perform when you just interrupt them every so often and go 'is that the best you can do?', or re-iterate instructions, or repeat the overall goal. I find instruction following _so poor_, especially for 'presentation layer' aspects. Yet if I ask the model to rewrite its last response, it does so perfectly. Why can't the model do this 'internally' and save me having to say 'try again'!
Just because the 'model' is autoregressive doesn't mean the system as a whole needs to present a single stream of immutable text.