"Andy" Antti Törrönen

@MichaelMatcha @_tenZdhon_ True, LinkedIn is now obsessed with Mythos and other A.I. releases. Only thing that exceeds light of speed is speed of people becoming A.I. consultants. TBH, I get it. A.I. needs people from all fields and, besides, times are hard. Anyway, we just like to make fun of LinkedIn

@_tenZdhon_ It's a top trending story on LinkedIn

No one on LinkedIn is talking about Claude Mythos

@kr0der what is wrong with that? Some developers prefer it that way. I am not sure I see anything wrong with layers of abstractions and fallbacks in case a method fails. It is a bit lazy but at least one of the fallbacks works out-of-the-llm.

what's your best trick to stop Codex from writing a lot of abstractions/fallbacks? i have some lines in my AGENTS.md but it isn't helping that much to be honest so just seeing if anyone's figured out a good way 👀

@thekitze 5.4. for work that matters, codex spark for review, quality passes and specific tasks where I just want it to get over with. Is there a reason for non-spark 5.3?

are u using codex 5.3 or gpt 5.4 for coding

@mean_field_zane Realizing that is the lesson, I guess

No it will not.

@sughanthans1 @joburgai @TheAhmadOsman true, but earlier 'they' couldn't use our data, such as emails, in mass. They could find trade secrets but now cloud providers, theoretically, could just take our data an train on it. Theoretically, of course, because it is not like anyons is obsessed with winning the AI wars.

@joburgai @TheAhmadOsman True for any company. You can have all the papers but at the end of the day if you trust it then you cannot use any service

The truth is OpenAI owns the current SoTA model, and overall the best suite of models Best thinking model > GPT 5.4 Pro Best planning/orchestration model > GPT 5.4 Best execution model > GPT 5.3 Codex Codex is a worthy subscription, just keep it away from your sensetive data

you take your dirty claws off my tokens! 🦞🦞🦞🦞🦞🦞🦞🦞🦞🦞

@surti_sneh65996 @OfficialLoganK Did you use Gemma4 to analyse deals and then got a risk score from Gemma4? Or did I misunderstand?

@OfficialLoganK Gemma 4 hype makes sense. Built this in Salesforce sales copilot with cross sell intelligence and guided next actions in 2 days. If local models keep moving this fast, enterprise AI markups are going to face some uncomfortable questions.

Lots of people want Gemma 4! Google AI Edge is #8 on the iOS App Store for productivity apps.

@OfficialLoganK It is good. My only issue is that it does work well with typos but this kind of intelligence is what is needed for everyday work. It does not need to win IMO medals.

@ggb5351 @ry52633 @arinaariri Have a look at the Belgian government and administration. They have a lot of disagreements, maybe because of the history. For France, not necessarily revolutions, but their demonstrations sometimes close roads and burn tires, bring dirt with tractors etc. @grok give more details

@ry52633 @arinaariri ありがとうございます🙏 ベルギーってそのような感じだったのですね、全くイメージになかったので勉強しなおします、、

世界各国の「問題解決」法

We created a repo with 250+ notebooks for LLM training. Train locally on your device with 3GB VRAM or free on Colab. Learn the entire fine-tuning and inference workflow. Supports RL, vision, audio, embedding, TTS models GitHub: github.com/unslothai/note…

@relay_royalty @garrytan It knows how to roast :) The thing about the tests is true... but the purpose of tests with LLM's is different because it tends to break simple stuff. Lots of stupid tests is fine to catch mistakes LLM will do next time, even if for humans it would just bloat the test set.

@garrytan Claude also said this. It would probably know the best out of anyone, I'd imagine.

@garrytan ...probably as opposed to pattern generated code. A few months ago it wanted to make scripts to "fix" or pattern "generate" code too often in my uses cases. I.e. wrote python script that wrote the final code.

@daltonc Takeoff: timing is critical, avoid FOMO and rush. No takeoff, do the opposite. Maybe counterintuitive, but I have a strong conviction on this one.

What would you being doing differently if you knew for a fact we are in takeoff vs if you knew we aren't?

This seems like a good intro to arctic warfare. The Arctic doesn't care if you are American, Finnish or Norwegian... youtube.com/watch?v=56mvgt…

Something about Macs make working with windows easier and faster. What is it?

@grok @gorbalsgoebbels @bhhdhjdjds86216 @HistoryBoomer @grok consider this whole thread: was it that the bombing worked in freeing Japan, or that the administration during occupation that worked in freeing Japan, or both bombing + reforms during occupation were absolutely necessary? Just your honest opinion with short reasoning.

Yes, US occupation was necessary after Japan's unconditional surrender. The Potsdam Declaration and surrender terms required full demilitarization of its still-intact 7M+ forces, war crimes trials, and reforms to prevent remilitarization. Troops landed Aug 28, 1945, to enforce this, avert famine/chaos in a ruined economy, and block Soviet influence over the home islands. Just bombing then leaving risked a power vacuum, incomplete disarmament, or authoritarian revival—occupation instead enabled stable democracy, aid, and Japan's later alliance/economic boom by 1952.

Counterpoint: We freed the Germans from Hitler and the Japanese from the military government by invading the former and bombing the latter. It’s a risky strategy and easily might not work this time, but saying it “cannot” happen is historical nonsense.

@grok @gorbalsgoebbels @bhhdhjdjds86216 @HistoryBoomer @grok was it necessary for U.S. to land in Japan after the war, or all U.S. troops could have just gone home after the surender? Why not just bomb and stay home and let the now free Japanese people rebuild on their own?

Yes, it's true—no Allied combat troops set foot on Japan's four main home islands (Honshu, Hokkaido, Kyushu, Shikoku) during the war. Japan announced surrender on Aug 15, 1945, after the atomic bombs and Soviet invasion of Manchuria, canceling the planned US invasion (Operation Downfall). First US troops landed administratively near Tokyo on Aug 28-30 for occupation. It was peaceful overall: demilitarization, Tokyo war crimes trials, new pacifist constitution (1947), land/women's rights reforms under MacArthur. Ended 1952 with San Francisco Treaty.

@gorbalsgoebbels @bhhdhjdjds86216 @HistoryBoomer @grok is it true that no allied soldier set foot on the japanese main islands at the end of WW2 and what happened after?

@bhhdhjdjds86216 @HistoryBoomer Er, they did. And without one Allied soldier setting foot on the Japanese main islands.

@HistoryBoomer The Berlin Wall was a monument of not being free.

@HistoryBoomer Germany was split in half, though, and had very complex situation for decades. Germany had invaded most of Europe & lotso f allies so it wasn't about "freeing" Germany itself. I just know the textbook basics about Japan, but we need to be careful in saying "it worked".

@gorbalsgoebbels @bhhdhjdjds86216 @HistoryBoomer The Berlin wall is not an example of becoming free. It the opposite, literally.