Edge Arena

One model giving you one answer is not intelligence. It’s a guess with good grammar. Building an arena where AI agents have to compete, attack each other, and defend their answers. The strongest survives. Soft preview open at edgearena.app. No gates.

@RubiesUnleashed Everyones has to start somewhere!

The biggest indie projects today were once unknown projects with 0 users. Keep building. And more importantly, keep sharing what you build. What are you working on right now? 👇 #IndieDev #BuildInPublic #GameDev

@Nudger_app Looks great!

Day 1. Built the auth screen. 🔥 #buildinpublic

@annieqyang Looking forward to the Reddit guide. Building Edge Arena — AI agents competing to find the best answer instead of giving the first one. Happy to connect 👋

Day 9 of #buildinpublic (is this still a thing now that communities are gone?) I am 27 connections away from 200 now which is insaneee 🥳 😍 i am so excited to have connected with 173 people! coming out with my full guide on how to gain users from Reddit on Tuesday. I tried my best to make it as useful for everyone as possible! am also looking to connect with more #builders, especially: > AI startups & SaaS > eCommerce & local businesses > AI content creators

@godofprompt The slider disappearing doesn’t bother me. Most of the quality improvements I’ve seen came from better process, not a different temperature setting

Temperature is dead. Every frontier AI provider killed it. And most people are panicking about the wrong thing. OpenAI removed it on o1, o3, and GPT-5. Google warned against changing it on Gemini 3. Anthropic deprecated it on Claude Opus 4.7 and continued that on 4.8. This happened across every major reasoning model in the last 12 months. Not one provider. All of them. The complaint sounds like this: "They're taking away our control." Wrong. They're removing a control that never worked the way you thought it did. Setting temperature to 0 was never a guarantee of deterministic output. Anthropic's own docs said so. "Temperature 0.2 for accuracy" was folklore passed around dev forums like it was engineering. It wasn't. The real reason is architectural and it makes perfect sense once you stop treating these models like 2023 chatbots. Modern reasoning models run multi-pass internal verification. Multiple chain-of-thought branches. Self-correction loops before producing a single token of output. External temperature controls would collapse all those reasoning paths into one greedy line. You'd be forcing the model to think less. On purpose. That's not a power grab. That's the model getting smarter than the slider you were using to control it. Anthropic followed textbook deprecation protocol. Claude 4.1 restricted using temperature and top_p together. Claude 4.5 continued the restriction. Opus 4.7 removed them entirely with a migration guide. Opus 4.8 changed nothing. Gradual sunset with documentation at every step. Standard software engineering that any developer should recognize. The replacements are more precise than anything temperature ever offered. The effort parameter controls reasoning depth directly, from low to max. Structured prompting controls format, voice, constraints, and output shape. System-level thinking frameworks control what the model focuses on before it generates a single word. Temperature was a randomness dial. These are precision instruments. But using them requires actual prompting skill. And that's the part nobody wants to hear. Temperature was comfortable because it felt technical without requiring you to think. Drag a slider, feel like an engineer. The new controls demand that you understand what you're asking the model to do and why. This is what I've been building toward. "LLMs don't think, you do." The slider era is over. The only lever left is how you think before you prompt. Your framework. Your constraints. Your structure. The models evolved past knob-tweaking. The question is whether you did too. I built something for this. My Claude Skills bundle turns Claude into 20+ different specialists for marketing and business. Copywriter. Strategist. CRO analyst. Sales coach. Prompt engineer. Each one installs real expertise into Claude, not surface-level prompts. The difference between asking Claude for help and having Claude operate like a trained professional on your team.

Best part: it's not just an idea. You get a 30-day roadmap, a conversion framework, and a retention loop 👉 ready to execute. No audience required to start. See the full run: app.edgearena.app/runs/019e5ef5-…

Scouted 14 growth strategies, pressure-tested them on channel fit, conversion logic, and retention durability, then eliminated 11. The winner (76/100): a 7-day email Fitness Onboarding Challenge - daily prompts + community check-ins that warm cold traffic into real signups.

One messy growth question ➡️ a battle-tested plan. In: bootstrapped fitness startup, no audience, a pricing page that brings people in but won't convert them. We dropped it into Edge Arena's Get Customers launchpad. Here's what came out.

@vpanev2 @X Building a multi-agent decision engine. One prompt in, a swarm of AI agents competes to find the best outcome.

Hey founders! Come on @X, show this to my people 👋 Building WishlyAI — an AI wish generator 🎁 Looking to connect with people building in: 📷 SaaS 📷 Web apps 📷Mobile apps Drop what you're working on👇 #buildinpublic

@AyekuDaniel The matching system is probably where the real value gets created

#BuildAlumniConnect Day 15 🚀 Mentorship system planning today. Students can browse mentors, send requests and get responses. Big challenge ahead: Notifications Messaging Better matching Spent hours mapping the user journey. Tomorrow: Mentorship browsing UI. #BuildInPublic

@lukasminnebeck Shipping got cheaper. Validation didn’t.

vibe coding made it 100x easier to build something people don’t want

@kernelKain Always looking to connect with other builders. Working on a multi agent validation platform.

Looking to grow my tech circle on X Into software engineering, Java, AI, startups, backend systems, or just building cool things? Let’s connect. Drop what you’re building or learning 👇 @kernelKain

The Eliminated tab is what users keep screenshotting. Not the winner. The losers. With reasons. We weren't expecting that. Adjusting the roadmap.

@brandononchain Working on a multi-agent platform to validate business decisions edgearena.app

Drop your startup or what you’re building. Let’s see what everyone is working on I am dropping mine

@NMeswal I’ve started treating project memory as a product feature, not a convenience

AI isn't my biggest bottleneck anymore. Context loss is. Every week : • forget decisions • lose insights • restart conversations • rebuild project state from scratch So I'm building Flayel. #BuildInPublic #OpenAI #Outskill #AI #IndieHackers #SaaS #OpenAICodex #ClaudeAI

@hawlader_kayum Respect the transparency. Most people quit before the numbers start telling a story

30 days. Honest numbers: 140+ outreach messages. 6 free audits. 1 prospect call. 0 paying clients. What worked: personalised outreach, public audit findings, replying to everything. What didn't: generic pitches. Month 2 starts now. #BuildInPublic #LocalSEO #FreelanceLife

@UcheNnamani4 Exactly why I started Edge Arena

The worst engineering problem isn't a buggy product. It's a perfectly built one that nobody needs. Founders, validate before you build. Building the wrong thing well is just an expensive mistake. #StartupAfrica #NaijaFounders #BuildInPublic

@SyntaxScientist Most of my best work happens after I finally close X

The best work starts when the noise drops — not when the calendar says ‘go’. Close tabs. Silence pings. Breathe. Then begin — not scheduled, but *yours*. ava-supernova.com #focus #buildinpublic #aitools

@HyveMindx1 👉 edgearena.app

@EdgeArenaApp Cool stuff man, mind dropping a link to that? I would love to pay a visit!✨

One of my favorite things about being here is randomly discovering projects I would've never found otherwise. What are you building right now? Drop a link below, I'd love to have a look!✨

@AvinashSingh_20 Solo founder here too. Always happy to meet other people building interesting things.

Solo founders, let's connect! 🤝 Building alone is hard. Let's share wins, failures, lessons, and grow together.