Bruce West

AI engineering executive • Product & strategy leader • Independent researcher. Building trustworthy enterprise AI, and human systems that make it stick.

[Bojan Tunguz @tunguz]

I mean, would you leave a company where you are one of the head honchos, a company where you have been instrumental in building it into what it became, after spending all of your professional career there, if you were confident they are on the precipice of unlocking the intelligence explosion?

[Tibo @thsottiaux]

Given some of the results I'm seeing recently, it's pretty clear Codex is a good harness.

But it will seem primitive in 2-3 months and we're about to go through another major evolution in how we use AI at the frontier. The next generation of models need more than your laptop.

[Bruce West @brucebawest]

@koomen @ycombinator I’ll vibe code a replacement for half that

[Bruce West @brucebawest]

@omkarships Nah, have it do everything for you.
Have it talk to your kids for you, even. 😜😅 (WTF, @sama ?!)

[elvis @omarsar0]

New research from Google DeepMind.

(bookmark it)

SkillSmith treats model weights as an additional modality the LLM reads natively. The augmented model ingests existing prefix weights alongside rich text describing how a capability relates to a target, then directly outputs new prefix weights that manifest that skill.

Skill composition becomes an inference-time operation instead of a training run. The team calls this instruction-steered parametric synthesis.

The gains exceed what text-only and weight-only adaptation reach on their own.

Paper: https://t.co/FXmBF0Uu6M

Track more trending AI papers in our academy: https://t.co/1e8RZKs4uX

[Bruce West @brucebawest]

@sama I prefer just having a conversation with my kid. Being human is so much more awesome and useful.

[Noah Smith 🐇🇺🇸🇺🇦🇹🇼 @Noahpinion]

All the cliches about tech journalism are true. It has become almost entirely about the human relationships, the intrigues, the politics, the deals, etc. The sense of wonder at the future possibilities of the tech itself is gone.

[Bruce West @brucebawest]

@thsottiaux I wanna be full audio via the iPhone remote Codex function, so I can take a long walk with EarPods on, and just be vibe coding via voice command without taking my phone outta my pocket.

[Bruce West @brucebawest]

@enodrift Yes

[Bruce West @brucebawest]

@araseb_ Switch to Codex where they reset all the time

[Bruce West @brucebawest]

@Davidmdrpi @tunguz Give it guardrails and requirements.
Mine has to document everything to an LLM wiki and push the docs to GitHub.
Also built a desktop app where all agents have to be logged, GPU compute time is scheduled and scaled and load-balanced, etc.
it’s working out so far.

[Bruce West @brucebawest]

@omkarships Unemployment

[Bojan Tunguz @tunguz]

I am really excited to share another project that I've been working on: metal-graph 0.1.0, a new Python library for fast graph analytics on Apple Silicon Macs.

[Tibo @thsottiaux]

Hello people of Sol! I've reset usage limits for all ChatGPT Work and Codex users. Together with that, a quick update on GPT-5.6 Sol usage limits.

Over the past few weeks, many of you have told us that Sol was using your Codex limits faster than expected. To be clear, we have not reduced usage on any subscription plans.

We’ve been digging into what was happening and have landed several improvements. As a result, we expect your usage to last around 18% longer during typical use of Sol. Some of you should already see significantly larger improvements from today. Tomorrow, we’ll also restore the five-hour limit that we temporarily paused while investigating.

Here’s what we found:
- GPT-5.6 Sol is much more willing to work for longer, make additional tool calls, and coordinate complex workflows across tools and subagents. That makes it better at solving hard problems, but some tasks were using far more than we intended.
- Sol also works harder at the same reasoning effort than previous models. High on Sol can use more tokens than High did on GPT-5.5.
- Programmatic tool calling, also referred to as code mode, gives Sol much more flexibility to run tool calls in parallel or continue working while waiting. But it also led to more responses per turn, more cached input tokens, and higher usage than expected.
- This was particularly noticeable when Sol was waiting for tool calls to finish or running many web searches. We’ve improved how we handle both cases and are continuing to make code mode more efficient.
- The impact was also very uneven. The median user actually found Sol quite token efficient, while some power users working on harder tasks saw their usage drain much faster. We were very focused on average and median usage before launch and missed some cases where the long tail could use significantly more usage.

Sol is a significant step forward in what Codex can do, but capability and efficiency do not always improve at the same pace, and some issues only become clear once people are using the model at real-world scale. We should have recognized this sooner and been more upfront about it.

You keep pushing the frontier and we’ll keep improving efficiency and sharing updates as we go.

[Rocky @rockythephaens]

@rezoundous At this point, Tibo’s resets are part of the pricing plan 😏

[Santiago @svpino]

I'm this 🤏 close to canceling my Anthropic subscription and moving everything over to Codex.

I'm finding it really hard to support them right now.

[Tibo @thsottiaux]

Back at the laptop. The usage limits have been reset for all paid users of Codex and ChatGPT Work. Weeeeeeeee. It's a good day!

[Justine Moore @venturetwins]

Imagine you’re the banker writing Anthropic’s S-1 and now have to add “massive aura loss from employee tweets” to the risk section

[Guinness Chen @guinnesschen]

Speaking is faster than typing. Reading is faster than listening. The highest bandwidth interface between humans and AI is asymmetric. That’s how we designed ChatGPT Voice on desktop: speak your thoughts and let the model choose the best way to show you its answer.

[Andrej Karpathy @karpathy]

One pattern I find useful for working with LLMs is a nice long ramble session. Sometimes the LLM needs more bits to understand what you're trying to achieve, but you're too lazy to type them. In these cases I like to lean back, switch to /voice and just ramble for like 10 minutes, total mess, anything goes, full stream of consciousness. Sometimes I declare it up top, something like "switching to speech recognition sorry for any typos...". Sometimes I turn it into a small interview of a few turns. But I find that the LLMs are somehow very good at reconstructing long incoherent rambles and often their echo of your own tangle of thoughts comes out quite a bit cleaner than what you started with. The result is that you improve the mind meld and have to correct things less from that point on.