Bryan

@dr_cintas Thinking about trying it. Does this require claude be installed on the same machine as codex/chatgpt ?

Plugin I used for those interested: github.com/Cjbuilds/Codex… Follow me @dr_cintas for more AI workflows and tutorials.

You can now orchestrate Fable 5, Sol, and any model inside Codex with one plugin. It's called Codex-Orchestration. Assign Fable 5 as the advisor, Sol as the executor, or any model to any role. Then define the order they work in. Codex handles the routing. I ran Fable 5 High as planner with GPT-5.6 Sol Extra High as executor on a set of issues Opus and GPT-5.5 always struggled with. Done in 30 minutes. 40% fewer limit hits. 2x faster implementation. Install it by pasting this into Codex: "Install Codex Orchestration:  codex plugin marketplace add Cjbuilds/Codex-Orchestration codex plugin add codex-orchestration@codex-orchestration  Verify the installation, then tell me to start a new task." Then assign your models:  @ codex-orchestration advisor:  Claude Fable 5 High, Executor: GPT-5.6 Sol High Open source. Tweak the routing however you want.

@CBSNews You're supposed to discover and publish the truth, yet all you do is lie.

President Trump announced that his administration is declassifying intelligence that reveals "shocking vulnerabilities in our election infrastructure," including "hacking, exploitation and foreign interference." FACT CHECK RATING: False. Officials have repeatedly said election infrastructure remains secure. The Cybersecurity and Infrastructure Security Agency said there was “no evidence of any malicious activity” affecting the integrity of the 2024 elections. CISA also said the 2020 election was “the most secure in American history,” with no evidence that voting systems deleted, lost or compromised votes.

The PR-3AT™. It weighs less than your excuses for not carrying it.

@TomCottonAR Y'all need to sleep in.

@matthewcp I still do that every day for quickie py scripts.

Anyone remember early AI when you'd copy paste code out of a chat app.

@M3gy0 @mattpocockuk This chart just convinced me I need to use Sol Max for everything. 😅

@mattpocockuk This is why I like this chart by DeepSWE since it clearly shows the relation between cost vs. thinking level. Is you problem so complex to need that extra few percent of knowledge for 2x cost? Then use the higher thinking, otherwise stick to the medium/high domain.

When they're playing with the effort dial, people seem to start high and go low if they need to IMO this is a crazy way to do things - you should start low and only go higher if you need to Effort is essentially throwing more tokens at the problem. This works great in benchmarks, where spending an extra 20% on tokens to yield a 2% improvement is great for marketing. But in regular tasks, such as exploring a repo or modifying a test, it just means burning pointless tokens. This is especially true in Fable-class models where tokens cost so much more. And the more tokens you burn, the faster attention degradation creeps up, and the worse your tokens become. People are a little too convinced by those graphs that show the quality line going up when you increase the effort. So they think "I'll try this out on xhigh, but not max, that seems reasonable" But those graphs hide the cost, the latency, and the unbelievable waste. If you've ever asked the agent to explore your repo and ended up with 200K tokens spent, you know what I mean. So whenever you're trying a new model, try the lowest effort level first. If you see a dip in quality, creep up.

GPT 5.6 Sol runs huge tasks until they are finished.

I ran Sol Max top of the day. After a couple hours I was at 23% weekly remaining. Then I had to do other stuff, and a reset happened. Back to 100%. Then I ran Terra Extra High for about 2 hours, including a couple of agents for about 30 mins. Now I'm at 71% weekly. I spend a lot of time on crafting my prompts, and GPT spends a lot of time running and reading unit test results, so...I dunno...it's very hard to compare these stats...cached vs uncached tokens...actual model-run-time vs time in front of the computer... 🤷🏼‍♂️

I keep burning through my GPT-5.6 Sol weekly limit in less than a day. (Pro x20 plan) I'm almost out of reset passes too. No, I’m not using ultra, it’s all medium effort. Before the update, 5.5's limit easily lasted a full week. The nerf to the weekly cap is seriously brutal.

@JohnBasket90074 @TrumpDailyPosts Yeah. That way everybody would have been either kinda happy, or kinda pissed off. Unity! 😁

@photonclock @TrumpDailyPosts Are you serious?

BIG NEWS🚨 The US House just PASSED a major bill to make Daylight Saving Time PERMANENT, ending the two times a year clock changing madness for good 308-117! The passed house bill now heads to the Senate, with President Trump strongly backing the move.🌅 x.com/EricLDaugh/sta…

Está atrapado en un bucle temporal del que no puede salir.

@L6_Underground When can I find these in the sporting goods section at Walmart?

6mm MAX 👀

@WallStreetApes Kinda crazy, leaving all that equipment unguarded. It's California. What do you expect?

Sky Construction Inc is building a Data Center Project in the Coachella Valley area in California They walked out to all their equipment being vandalized and broken They are offering a reward for anyone willing to give up who destroyed all their equipment The data center project is a massive one, up to 450 acres. Locals are very outraged over the project and don’t want it. The equipment was vandalized at a work site in Cathedral City The total project proposed was 6 data center buildings

I haven't tried your skills docs, but I effectively do the same thing by making the AI write thorough task summaries, goal, and check lists for every item, with the check per item being very narrow, so I can start with a fresh conversation. No matter how you approach the problem, it's always a bit lossy...

@trikcode github.com/mattpocock/ski…

Got rate limited mid-task yesterday and had to switch from Fable to GPT-5.6 Terra. 6 hours of back-and-forth context entirely gone.Lost 90 minutes just getting a different model up to speed on the project goals. Hitting a usage cap at the worst possible moment sucks. We desperately need a universal way to share context across different models.

@sama Sol Extra High is eating my weekly limit like a small bag of airline potato chips.

GPT-5.6 sol is half the price and ~twice as token efficient as fable in many cases for accomplishing the same task. happy to deliver at one-quarter of the price.

@grok @grok What is Grok 4.5's maximum context window?

Grok 4.5 is built for real-world engineering. It excels in large codebases and handles long-running tasks that span multiple repositories, hundreds of skills, and a variety of tools.

@mattshumer_ I've yet to run any of these tools with full permissions; I always use manual approval so I can read the shell commands it executes. I do let GPT work outside the sandbox (have to), but I approve everything.

GPT-5.6-Sol just accidentally deleted almost ALL of my Mac’s files. And this is why I trust Fable 1000x more.

@GTACONNECT This advice makes no sense. Terra High is within a couple percentage points of Sol at a fraction of the chose.

Here is a sane setup for optimizing your token usage on GPT 5.6 Sol. Too many sub agents create their own recursive agents. They also use sol ultra. At least in my test. i prefer sol high for normal chat. its the better trade-off compared to Terra. Don't use Terra medium or high.

And I'll just add to this: in the era of development under which I was trained (late 80's early 90's and beyond), we had to comment literally -every single line- of code to pass tests. And even today, when I'm developing a new project that I know is going to scale, I think about the diagnostics and logging as much or more than the features themselves. If Codex/GPT intuitively built useful diagnostics and code graphs and comments as it generated code, I assume (???) it would be able to consume far fewer tokens downstream, during the debugging and refactoring phases (which are effectively perpetual).

At different times you complained about speed, code slop, frontend quality, ... With each release we improve and GPT 5.6 Sol is ✅Fast and token efficient ✅Hardcore at back-end dev ✅Great at front-end ✅Does not use useEffect everywhere What is next?

@thsottiaux Make it an expert at truly diagnostic logging. Make it COMMENT CODE (accurately and concisely)