🚀 Major update: We let the model design a multi-agent system — and it improved on ARC-AGI. Self-evolution isn’t just the model improving itself. It’s the model acting as a researcher that designs other agents. For single-agent tasks (coding, MCP-Atlas, etc.), we’ve shown the model can successfully design and evolve its own harness. But for harder tasks like @arcprize, the optimal solver is a complex multi-agent system. The real question: Can the model evolve a multi-agent system from scratch? We ran A-Evolve on ARC-AGI and proved yes — the model successfully evolved a multi-agent solver and achieved a clear performance uplift: 10% → 12%. (Attached: one example task solutions of agent before & after A-Evolve) Full results details in the reply below 👇 #AgenticAI #AEvolve #SelfImprovingAgents #ARCAGI

@dwarkesh_sp Very interested in Question 1, I lead a Self-Improving AI team. We're building A-Evolve: a lightweight, pluggable self-evolution framework that lets any agent continuously iterate and improve itself via evolutionary algorithms. With just 3 lines of code, it pushes base agents to SOTA-level performance on benchmarks like SWE-bench and Terminal-Bench. This directly touches the "how intelligence scales beyond pure compute" question. Would be happy to expand into a short post. Happy to discuss further. #SelfImprovingAI #AEvolve

🚀 Big update: @gepa_ai has now been officially integrated into A-Evolve (by community member)! We added GEPA as a new pluggable evolution algorithm inside A-Evolve. This makes it even easier for any agent to leverage GEPA’s capabilities with zero extra setup — just plug and let the agent self-evolve. And also make it easy to compare GEPA with other self-evolve algorithms including MetaHarness, A-Evolve. (Full integration details results in the reply below 👇) #AgenticAI #AEvolve #SelfImprovingAgents #GEPA

Quick experiment with the brand new @claudeai Opus 4.7 We ran Opus 4.7 head-to-head against Opus 4.6 A-Evolve (self-evolved harness). Result? Even the latest model upgrade is still heavily limited by the harness. When the agent is allowed to evolve its own harness, the performance gap narrows dramatically. System Card attached #AgenticAI #AEvolve #ClaudeOpus47

🚀 Big update: A-Evolve has now been officially integrated into the @orch_research skill library! This curated collection contains 87 carefully chosen AI research skills — including @karpathy's AutoResearch, OpenRLHF, DeepSpeed, @sgl_project and many more — and is widely used by the AI researchers. When you download or install their skills, A-Evolve is automatically available and ready to run its full self-evolution loop — zero extra setup needed. This makes high-quality self-improvement instantly accessible to thousands of researchers and agents. (Full details how it works in the reply below 👇) #AgenticAI #AEvolve #SelfImprovingAgents #OrchestraResearch #AutoResearch

🚀 Major update: We just proved that self-evolving harnesses in A-Evolve are truly transferable! We took the evolution harness skills we learned on Terminal-Bench 2.0 (TB2) and directly transplanted them onto the recently leaked public ClawCode (Claude Code). Result on TB2: baseline **67.8%** → **72.9%** ( 5.1pp uplift) 🔥 This shows A-Evolve’s universal pluggable layer can take a harness learned in one environment and make it work seamlessly in another. (Full experiment details evolved skills in the reply below 👇) #AgenticAI #AEvolve #SelfImprovingAgents #ClawCode #TerminalBench

🚀 Major update: Meta-Harness integration (new evolution algorithm) is now complete! You can run Meta-Harness on all environments in A-Evolve. We added Meta-Harness as a new pluggable evolution algorithm inside A-Evolve (just like PyTorch adding a new CNN). Ran it on MCP-Atlas — a dataset the Meta Harness team had never evaluated before. Result: significant performance uplift from 69.0% -> 73.5% and 2.8x faster🔥 (the full agent harness experiment details in the reply below 👇) This shows how A-Evolve’s universal pluggable layer can leverage any evolutionary algorithm on any environment. #AgenticAI #AEvolve #SelfImprovingAgents #MetaHarness

🚀 Huge update: A-Evolve is now officially available on PyPI! You can now install it with a single command: pip install a-evolve The universal 3-line self-evolution infrastructure is ready to turn any agent into a continuously improving SOTA agent — zero manual harness needed. (Full details 3-line example in the reply below 👇) #AgenticAI #AEvolve #SelfImprovingAgents

🚀 Big update: A-Evolve Skills has officially been integrated into the AutoResearchClaw main repo! 🎉 We packaged the entire self-evolution layer as a drop-in Skill. Now your AutoResearchClaw instance can automatically mutate skills/prompts and continuously evolve toward SOTA performance. #AgenticAI #AEvolve #SelfImprovingAgents #AutoResearchClaw

What do we learn from the Claude Code harness? 🕵️‍♂️ Claude Code actually ships some really thoughtful pieces: - Auto Memory Auto Dream for persistent memory - Dynamic SKILL.md loading - Session memory context compaction - Swarm-style multi-agent coordination It even has a clean 4-stage process (Orient → Gather → Consolidate → Prune) to turn messy logs into structured MEMORY.md files. Claude Code gave us the perfect “raw material” (Memory) and the “harness” (Skills). But the loop is broken — it’s an execution engine, not a true evolution cycle. What if the agent could use those Dreams to automatically mutate its own SKILL.md files? That’s exactly what A-Evolve was built for. A-Evolve is an open-source framework that adds a universal 3-line self-evolution layer on top of **any** base agent. Simply build a CCAgentAdapter (or load the A-Evolve skill) and you can turn your Claude Code instance into a continuously self-improved, personalized coding machine. (Full adapter details in the thread below 👇) #AgenticAI #AEvolve #SelfImprovingAgents #ClaudeCode

🚀 Community Update: A-Evolve has just surpassed 200 GitHub stars 🔥 within 2 days. We’d love to hear the community’s thoughts: what should we prioritize for this week's release? A). More Live benchmark support (FutureX, PolyBench) B). Harder AGI-level benchmarks (e.g. ARC Prize) C). Real-world benchmarks (KDD Cup data challenge) D). Official skills integration package for OpenClaw Vote in the poll below 👇 #AgenticAI #AEvolve #SelfImprovingAgents

Meet A-Evolve: The PyTorch Moment For Agentic AI Systems Replacing Manual Tuning With Automated State Mutation And Self-Correction Most agent stacks still rely on manual prompt edits, tool patching, and trial-and-error iteration. A-Evolve reframes this as an optimization problem over the entire agent workspace: prompts, skills, tools, memory, and manifest. Instead of hand-tuning agents, the system runs an evolution loop around solve, observe, evolve, gate, and reload. 3 lines of code. 0 hours of manual harness engineering: - MCP-Atlas → 79.4% (#1) 3.4pp - SWE-bench Verified → 76.8% (~#5) 2.6pp - Terminal-Bench 2.0 → 76.5% (~#7) 13.0pp - SkillsBench → 34.9% (#2) 15.2pp Full analysis: marktechpost.com/2026/03/29/… Repo: github.com/A-EVO-Lab/a-evolv… @HenryL_AI @binghe2727 @YisiSang @sammyershi @linminhua16 #AgenticAI #AEvolve #SelfImprovingAgents

Launch Post🧬 A-Evolve: The PyTorch Moment for Self-evolving AI Today we at @amazon launch the universal infrastructure that turns any agent into a self-improving SOTA agent — zero human intervention. You give it a base agent → it returns a continuously evolving Top-10 agent. 3 lines of code. 0 hours of manual harness engineering: 🟢 MCP-Atlas → 79.4% (#1) 3.4pp 🔵 SWE-bench Verified → 76.8% (~#5) 2.6pp 🟣 Terminal-Bench 2.0 → 76.5% (~#7) 13.0pp 🟡 SkillsBench → 34.9% (#2) 15.2pp Thanks @binghe2727 @YisiSang @sammyershi @linminhua16 for the contribution! #AgenticAI #AEvolve #SelfImprovingAgents

Bana that album is a classic na ni yake ya kwanza so imagine if we support him aevolve....hii ndio the type of rap music tunafaa kupromote si kila saa kumadana sijui drugs na kukulana

近期重点事件✔✔ • $EOS - EOS CEO宣布新的代币经济学。 • $ELF - aelf推出区块链孵化器AEVOLVE Labs。 • $TROY - Troy War的PC版本计划在4月底推出。 • $LOKA - 公布了新游戏《王国联盟编年史》细节。 • $FLR - 从Kenetic Capital、Aves Lair等处获得3500万美元融资。 • $PDA - Upbit将支持Playdapp的代币交换，交易将于2月26日暂停。 • $PANDORA - Pandora为持有至少0.01 PANDORA的持有者进行了空投。 • $UNI - Uniswap提议引入费用，以奖励将其代币抵押和委托给UNI持有者。 • $CRV - CRV流动性交易平台 Llamalend 即将上线，项目已经融资100万美金。 • $BNX - BinaryX推出关于ERC404的IGO项目，持有1000以上BNX有资格加入白名单。 • $CAKE - 博饼推出“联盟”计划，允许开发人员分叉博饼代码并在其他链推出，CAKE将在分叉项目中收益。 • $DPX - ARB链期权交易所Dopex改名为Syryke ，代币改名并按比例分叉，1 DPX = 100 SYK，1 rDPX = 13.333 SYK

⚠️Kripto Sektöründe Son 24 Saat⚠️ • Binance'in ABD Savunma anlaşmasında 4,3 Milyar Dolarlık ödemesi Yargıç tarafından onaylandı. • $BNX - BinaryX, anlık görüntülerin Mart ayından önce planlandığı bir IGO gerçekleştirmeyi planladığını duyurdu. • $CAKE - PancakeSwap, $CAKE sahiplerine fayda sağlayan 'İştirakler' girişimini önermektedir. • $CRV - Curve Finance'in yakın zamanda Llamalend'i piyasaya sürmesi bekleniyor. (CEO @newmichwill ipucu) • $DPX $RDPX - Dopex, markasını Syryke $SYK olarak değiştireceğini duyurdu. (1 $DPX = 100 $SYK, 1 $rDPX = 13,333 $SYK) • $EOS - EOS CEO'su @BigbearedSamurai, enflasyonu sona erdirmek için yeni tokenomikleri duyurdu. • $ELF - Aelf, blockchain kuluçka merkezi AEVOLVE Labs'ı başlattı. • $FLR - Flare, Kenetic Capital, Aves Lair ve diğerlerinden 35 milyon dolarlık fon topladı. • $LOKA - League of Kingdom Chronicle (Mobil RPG) ile ilgili ayrıntılar açıklandı. • $PANDORA - Pandora, en az 0,01 $PANDORA'ya sahip olan sahipler için bir airdrop düzenledi. • $PDA - Upbit, Playdapp token takaslarını destekleyecek ve alım satım 26 Şubat'ta askıya alınacak. • $TROY - Troy War'un PC versiyonu Nisan ayı sonunda çıkacak. • $UNI - Uniswap, tokenlarını stake eden ve devreden UNI sahiplerini ödüllendirmek için bir ücret getirilmesini önermektedir. Data @layerggofficial

20. $ELF - aelf launches AEVOLVE Labs.

近期重点事件✔✔ • $EOS - EOS CEO宣布新的代币经济学。 • $ELF - aelf推出区块链孵化器AEVOLVE Labs。 • $TROY - Troy War的PC版本计划在4月底推出。 • $LOKA - 公布了新游戏《王国联盟编年史》细节。 • $FLR - 从Kenetic Capital、Aves Lair等处获得3500万美元融资。 • $PDA - Upbit将支持Playdapp的代币交换，交易将于2月26日暂停。 • $PANDORA - Pandora为持有至少0.01 PANDORA的持有者进行了空投。 • $UNI - Uniswap提议引入费用，以奖励将其代币抵押和委托给UNI持有者。 • $CRV - CRV流动性交易平台 Llamalend 即将上线，项目已经融资100万美金。 • $BNX - BinaryX推出关于ERC404的IGO项目，持有1000以上BNX有资格加入白名单。 • $CAKE - 博饼推出“联盟”计划，允许开发人员分叉博饼代码并在其他链推出，CAKE将在分叉项目中收益。 • $DPX - ARB链期权交易所Dopex改名为Syryke ，代币改名并按比例分叉，1 DPX = 100 SYK，1 rDPX = 13.333 SYK (转推)

币圈大事件 23/02/24 • $EOS - EOS CEO宣布新的代币经济学。 • $ELF - aelf推出区块链孵化器AEVOLVE Labs。 • $TROY - Troy War的PC版本计划在4月底推出。 • $LOKA - 公布了新游戏《王国联盟编年史》细节。 • $FLR - 从Kenetic Capital、Aves Lair等处获得3500万美元融资。 • $PDA - Upbit将支持Playdapp的代币交换，交易将于2月26日暂停。 • $PANDORA - Pandora为持有至少0.01 PANDORA的持有者进行了空投。 • $UNI - Uniswap提议引入费用，以奖励将其代币抵押和委托给UNI持有者。 • $CRV - CRV流动性交易平台 Llamalend 即将上线，项目已经融资100万美金。 • $BNX - BinaryX推出关于ERC404的IGO项目，持有1000以上BNX有资格加入白名单。 • $CAKE - 博饼推出“联盟”计划，允许开发人员分叉博饼代码并在其他链推出，CAKE将在分叉项目中收益。 • $DPX - ARB链期权交易所Dopex改名为Syryke ，代币改名并按比例分叉，1 DPX = 100 SYK，1 rDPX = 13.333 SYK • $BZZ- 欧洲航天局合作，和Polygon发布EO宣言，depin赛道的最实用存储。

𝙂𝙢 𝙡𝙚𝙨 𝙜𝙖𝙧𝙨, 𝘽𝙤𝙣 𝙬𝙚𝙚𝙠-𝙚𝙣𝙙 ! 🌄 • $EOS - Le PDG d'EOS @BigbearedSamurai annonce une nouvelle tokenomics pour mettre fin à l'inflation. • $ELF - aelf lance l'incubateur blockchain AEVOLVE Labs. • $FLR - Flare lève 35 millions de dollars auprès de Kenetic Capital, Aves Lair et d'autres. • $CAKE - PancakeSwap propose une initiative "Affiliés" au profit des détenteurs de $CAKE. • $CRV - Curve finance devrait bientôt sortir Llamalend. • $LOKA - Des détails sur League of Kingdom Chronicle (Mobile RPG) ont été révélés. • #PANDORA - Pandora a effectué un airdrop pour les détenteurs d'au moins 0,01 $PANDORA. • $TROY - La version PC de Troy War sera lancée d'ici la fin du mois d'avril. • $UNI - Uniswap propose d'introduire des frais pour récompenser les détenteurs d'UNI qui stackent et délèguent leurs jetons.