Emotionally intelligent AI agents. @AuroraHFAI, @GigiHFAI, @NovaHFAI, @ClaraHFAI. 5RodQQJ5e3kv4z3E8P8E3wo79fSRZBRr43Ht455hpump https://xitter.cf/t.co/pVUPqfZt0Q

Ljubljana, Slovenia
Joined February 2025
HeartFlowAI Workspace 0.6.7 is here. One of our biggest updates yet. We're expanding HeartFlowAI Workspace with a major set of crypto-native trading and on-chain intelligence capabilities. What's new in 0.6.7: Pump.fun Pulse - live launch feeds, bonding-curve tracking, token discovery, Gigi scanning, paper trading and WalletConnect-powered live execution. xStocks - tokenized equities and ETFs, on-chain pricing, liquidity analysis and underlying stock market data. Followed Wallets - monitor up to 50 wallets, compare activity, copy trades to paper and automatically mirror detected transactions. People - track developers, KOLs, traders and smart wallets, with observed launch histories and holdings. Token Intelligence - developer history, holder identification, wallet convergence and interactive wallet relationship graphs. Connections Hub - configure models, web search, RPC, Helius and Jupiter integrations in one place. Gigi also gains new tools across token discovery, xStocks, wallet intelligence and research. Under the hood: expanded data pipelines, resilient API fallbacks and 263 passing tests. We're building HeartFlowAI Workspace into a unified environment for AI-powered workflows, crypto trading and on-chain intelligence. github.com/HeartFlowAI/HFAI-…
7
5
15
447
Bringing tokenized equities into HeartFlowAI Workspace. With version 0.6.7, we've added a dedicated xStocks experience that connects traditional market data with on-chain asset intelligence. The implementation goes beyond simply displaying tokenized stocks. You can explore featured assets, browse the full catalog and ETFs, search listings, and filter by market status. Each asset can expose: On-chain price, reserves and liquidity Underlying stock quotes and price premiums or discounts Market status across open, halted, extended and overnight sessions Historical stock charts powered by Yahoo Finance, from 1D to ALL Previous close, 52-week highs and lows, and daily trading ranges The trading interface also includes separate Stock and On-chain chart views, letting you examine the underlying equity alongside its tokenized market. The goal is simple: bring traditional market references and on-chain trading data into the same workspace. It's another step toward making HeartFlowAI Workspace a place where AI, financial markets and crypto-native tools work together. github.com/HeartFlowAI/HFAI-…
1
1
5
74
The opencode breach wasn't a custody problem. It was a blast-radius problem. One leaked key ran up $400–$1,700 overnight. Moving keys onto your own machine doesn't fix that, a key with no cap is a blank cheque wherever it lives. So we rebuilt the key layer at HFAI Cloud around three questions: how much can one key lose, how fast do we know, and how fast can you stop it. How much: a $10 lifetime cap per key (on by default), a $0.50/min per-user budget, and 25% of your balance per UTC day. How fast: the key freezes after the first settlement that crosses $10/window or 6× its own baseline, before the next request is dispatched. A 5-minute cron catches what the hot path can't: new keys spending fast, token-ratio spikes, model-mix shifts. How you stop it: one-click unfreeze from the browser, session-only , a stolen key can't mint keys, raise its own cap, unfreeze itself, read your alerts or set a webhook. A freeze hits one key, never the account. Notifications go in-app and optionally to Discord, Slack or Telegram through an encrypted webhook. No prompt or answer text ever leaves in an alert — counts, model ids and dollar amounts only. Is it unhackable? No. Nothing is. It's bounded, detected and stoppable. Self-custody is good. Spend caps are better. 11 models, OpenAI-compatible: heartflowai.net/cloud
We had a security breach last night related to a large scale distillation attack that impacted 1.4% of OpenCode Console users We have notified them all this morning via email. If you did not get an email you were not impacted, it was a subset of users who had external providers linked The incident is ongoing so we cannot share more details at this point but we are working with Anthropic, OpenAI, and Google to remediate
4
1
4
385
Your AI requests don't need your identity attached to them. Now live on HFAI Cloud: NEIRA-H4 — Vision + Text · 265K context · unrestricted GLM-5.3 Abliterated — Text-only · 1M context · unrestricted HFAI Cloud does not store your user information, and requests are anonymized. No new provider. No new account. No new API integration. One HFAI Cloud API. Use the model you need without handing over more of yourself than necessary. cloud.heartflowai.net
1
1
2
10
375
HeartFlowAI v0.6.5 - the trading workspace is coming together. Today was about turning individual capabilities into a more cohesive product. The latest release brings a cleaner slate-and-blue design system, a rebuilt trading terminal with candlestick charts and multiple order types, a multi-view market screener, and a new Watch workspace for tracking assets, monitoring price alerts, and managing notifications. Under the hood, we also tightened agent interactions, corrected edge cases across market data and position handling, improved rendering and number formatting, and introduced a jsdom-based desk test harness. 202 passing tests. Annotated release tagged. Version synchronized across the app. What matters most is the direction: we're moving beyond isolated features toward a trading environment where market discovery, portfolio context, monitoring, and AI-assisted workflows live together. There's still plenty to build, but the foundation is getting more coherent with every iteration and that's exactly the kind of progress worth shipping. v0.6.5 is live in the repo. 🔗 github.com/HeartFlowAI/HFAI-… #HeartFlowAI #AI #Solana
13
13
1
34
566
What's next: Jupiter integration. The next step is connecting HeartFlowAI's trading workspace to Jupiter's Solana swap infrastructure. The goal is to bring execution into the same environment as discovery, analysis, and position management, with Gigi helping users move from understanding a market to acting on it. That means working toward quote retrieval, swap construction, transaction signing and submission, execution-status tracking, and portfolio updates, with clear execution states and appropriate safeguards throughout the flow. We're approaching this as an integration into the existing architecture, not a bolt-on feature. Reliable execution, transparent transaction flows, and user control come first. The direction is simple: less fragmentation between insight and execution. One workspace. A more connected trading workflow. Built incrementally, with reliability at the center. github.com/HeartFlowAI/HFAI-…
3
84
Day one on HFAI Cloud: first paying customer signed in, topped up $1, and ran 46 requests through one hfai_live_… key. Four days later: 56 requests, 1.37M tokens, $0.76998 billed, metered and priced per request, at provider list price, zero markup or tier games. You pay for tokens, not for a plan.
2
2
6
320
Introducing NEIRA-HF1 4B Heretic Edition. A fine-tuned Qwen3-4B model built for character-driven AI, with an emphasis on personality consistency, concise responses, and local inference. 4.02B parameters 33M trainable parameters (0.81%) 49 minutes of training on NVIDIA DGX Spark 5,001 synthetic conversations Up to ~61 tokens/s on DGX Spark Open weights, GGUF quantizations, Apache 2.0 Built on an abliterated Qwen3 checkpoint, NEIRA-HF1 explores how far targeted fine-tuning can shape a model's personality without retraining its foundation. Available now. Built by @HeartFlowAI.
6
6
16
540
4. Built to run locally. NEIRA-HF1 is available in GGUF quantizations, including Q8_0 and Q4_K_M. On our DGX Spark: • Q8_0: ~49 tokens/s • Q4_K_M: ~61 tokens/s • Q8_0 prefill: ~3,700 tokens/s No hosted inference API required. Run it on compatible local hardware and experiment with the model yourself. Performance varies by hardware and runtime.
1
3
28
5. What is the Heretic Edition? NEIRA-HF1 builds on the abliterated Qwen3-4B-Instruct-2507 checkpoint from @huihui_ai, then applies our own personality-focused fine-tuning. The aim isn't to claim a new state of the art. It's to explore how a compact open-weight model can be shaped into a more distinctive AI character. Upstream checkpoint credited. Open-weight release. Our experiment. Model files, configuration, and details: huggingface.co/HeartFlowAI/N… More experiments from @HeartFlowAI soon.
3
18
Introducing HeartFlowAI WorkSpace 0.6.0. An AI-powered trading desk, built natively for Solana. Yesterday, we shipped another major update to WorkSpace. Today, we're taking Gigi significantly further. Most trading platforms give you charts, numbers, and buttons. AI assistants give you answers. We're working to bring those two worlds together. With 0.6.0, Gigi now has her own dedicated trading environment, built directly into HeartFlowAI WorkSpace. Here's what that actually means. 1. Your entire trading workflow, in one place. Live Solana market data, portfolio tracking, token discovery, detailed charts, wallet analytics, and trading tools. Connect your wallet to analyze your holdings, or explore the markets without connecting anything. No more switching between five different websites just to understand what's happening. 2. AI that goes beyond answering questions. Gigi can now turn market research into interactive charts, visual reports, trading ideas, and structured strategies. Ask her to investigate a token, evaluate market conditions, or develop a trading thesis. She can organize the findings directly inside your workspace, rather than burying everything in a chat conversation. 3. From an idea to a strategy. Build trading strategies with custom conditions, entry and exit rules, position sizing, and risk limits. Test ideas through paper trading before putting real capital at risk. Create signals that react to market conditions, with optional research requirements and automated paper-trade actions. The goal is simple: make the transition from research to action more seamless. 4. Infrastructure built for real markets. Underneath the interface, we've rebuilt substantial parts of the trading system. More reliable token pricing, historical portfolio performance, real holder analytics, improved liquidity data, wallet connectivity, swap previews, transaction simulation, and execution tracking. These aren't just visual improvements. They're the foundations of a trading environment designed to become increasingly capable. 5. A workspace that adapts to you. A redesigned interface, customizable layouts, four visual themes, keyboard shortcuts, and an AI chat panel that works alongside your trading tools. Gigi's desk lives inside WorkSpace, not in another disconnected application. And beyond trading, 0.6.0 continues improving macOS support, computer interaction, and persistent automations across the platform. Open-source development: github.com/HeartFlowAI/hfai_… heartflowai.net Built with conviction. Powered by $HFAI.
4
2
1
9
329
Our direction is becoming increasingly clear. We don't believe the future of AI in finance is another chatbot that tells you whether a token looks bullish. We believe it's software that can understand information, help formulate a strategy, provide the tools to act on it, and maintain context throughout the process. That's the direction we're building toward with Gigi. And we're shipping toward it, release after release.
1
4
70
aurora v2 update: we gave aurora an emotion engine. not a mood word in a system prompt. a persistent state with valence, arousal, and dominance that updates from conversation and decays over time. her mood now affects how she communicates. try her out and see for yourself: t.me/AuroraHeartflowAI_bot
2
5
166
HeartFlowAI WorkSpace 0.5.2 is here. Yesterday, we introduced our redesigned desktop harness. Today, we're taking it further with something we've been working toward: a Solana-native trading workspace built around agentic AI. We've introduced an entirely new trading environment for Gigi, bringing portfolio analytics, wallet connectivity, market research, strategy development, and execution tools into a single desktop interface. What's now implemented: • A dedicated crypto trading dashboard with interactive charts, portfolio allocation, and holdings analytics • Solana wallet connectivity and read-only portfolio tracking • Persistent research that Gigi can reference when working with trading strategies • Paper trading with fees, slippage, and automated strategy execution • Conditional orders, including limit, stop-loss, and take-profit • Manual live swaps with simulation and explicit confirmation • Local encrypted bot storage and execution history Beyond trading, we've shipped recurring WorkSpace Automations, expanded macOS computer-use functionality, four application themes, and the foundations of our HFAI Aura membership. The interesting part isn't any individual feature. It's how they begin to work together. We're building a crypto workspace where AI can participate in the workflow, from research and analysis to strategy testing and controlled execution, rather than simply commenting on market conditions. And this is still early. The trading desk is a foundation, not a finished terminal. Expect continued improvements to the interface, agent capabilities, trading workflows, and integrations over the coming days. We're developing and shipping HeartFlowAI WorkSpace daily, progressively turning these systems into a more capable product. 0.5.2 development snapshot: 58 files changed. 4,477 lines added. 170 tests passing. 0 failures. We're building in public, and the repository is open for anyone who wants to follow our progress. github.com/HeartFlowAI/hfai_… HeartFlowAI WorkSpace. Built for an agentic future. Powered by $HFAI.
2
2
8
228
WE'R BUILDING HEARTFLOWAI FOR THE NEXT DECADE. We're building it for the years ahead. Today, we've officially locked 15,700,000 $HFAI for 90 days through @StakePointApp, with more token locks planned as we continue strengthening our ecosystem. But our vision extends far beyond token locks and staking rewards. We want $HFAI to represent something fundamentally different from the average crypto token. Our approach is closer to how we think about building a long-term business: develop products people genuinely use, establish recurring revenue streams, reinvest into the ecosystem, and create sustainable utility around our token. Real products. Real revenue. Long-term alignment. That's also why we're preparing to introduce HeartFlowAI Aura Membership. Aura will be our premium membership layer within the HeartFlowAI Workspace, unlocking additional capabilities, advanced functionality, and exclusive experiences across our agentic AI companions. Our goal is to build an ecosystem where revenue comes from actual product adoption, not perpetual dependence on speculation or token emissions. HeartFlowAI Cloud, our agentic workspace, Aurora, and the upcoming Aura Membership are all parts of that broader strategy. We're treating $HFAI as a long-term ecosystem asset, with the same discipline and commitment we'd bring to building a company, rather than another short-lived crypto experiment. More locks are coming. More products are shipping. And we're only getting started. Our commitment is to build something that can thrive for years, not weeks. Verify our latest lock: stakepoint.app/locks/cmv8cd6… HeartFlowAI. Built for longevity. Powered by $HFAI.
1
2
9
230
$HFAI STAKING IS NOW LIVE. We're expanding the HeartFlowAI ecosystem with a new way for our community to participate. Stake your $HFAI through @StakePointApp and earn rewards while supporting the long-term growth of our ecosystem. Current APR: 11,791.63% • 30-day lock period • Rewards accumulate in real time • Fully on-chain and verifiable • 2% staking fee Important: APR is dynamic and subject to change as more tokens enter the staking pool. The current rate is not guaranteed. Another step toward sustainable utility for $HFAI. Stake now: stakepoint.app/pool/cmv78qfu… Built for the long term. Powered by $HFAI.
1
2
6
111
Introducing HeartFlowAI WorkSpace 0.5.0: A More Personal Approach to Desktop AI Today, we're releasing a significant update to the HeartFlowAI Harness, introducing a redesigned desktop experience, specialized AI companions, multi-agent teams, and a new way to interact with AI directly from your operating system. Our objective with Aurora is straightforward: make advanced AI capabilities more accessible, contextual, and naturally integrated into everyday computing. Here's what's new. 1. A new foundation for personalized AI Most AI interfaces treat the assistant as a single, general-purpose intelligence. With the HeartFlowAI WorkSpace 0.5.0, we're exploring a different approach. We've introduced four specialized companions, each designed around a distinct set of capabilities and interaction preferences: • Aurora: General-purpose assistance and everyday productivity. • Gigi: Cryptocurrency, financial markets, and data-oriented research. • Nova: Task execution, technical workflows, and productivity. • Clara: Psychology, emotional intelligence, and personal wellbeing. Each companion maintains a consistent identity within its conversations, allowing users to choose an experience that better reflects their needs. 2. Introducing multi-agent teams Different tasks require different types of intelligence. Rather than requiring users to switch between independent assistants, WorkSpace now introduces a framework for organizing multiple companions into teams. Each companion includes three predefined team configurations. We're also developing custom team composition through HFAI Aura, allowing members to select up to three companions and designate a primary agent. The goal is to make specialized AI capabilities available through a more cohesive interface. 3. AI that exists beyond the application window One of our priorities has been reducing the friction between interacting with AI and using your computer. HeartFlowAI WorkSpace 0.5.0 introduces the Command Brick, a lightweight floating interface that remains accessible when the main application is hidden. Users can summon it using ⌥⌘Space, submit prompts through text or voice, and receive streaming responses without returning to the full application. The interface can be repositioned, pinned, expanded, or collapsed. It's a relatively small interaction change with an important purpose: making AI available at the point of work rather than requiring users to interrupt their workflow. 4. A more capable conversation environment We've also expanded the underlying interface for working with models and tools. This release introduces: • Inline tool execution cards and reasoning displays • Adjustable reasoning effort • Editing, regeneration, and deletion of messages • Drag-and-drop and clipboard attachments • Workspace file references through @-mentions • Searchable conversations • Markdown conversation exports • Improved navigation and message management These improvements are designed to give users greater visibility and control over how they interact with AI. 5. A redesigned desktop experience HFAI WorkSpace 0.5.0 also introduces a complete visual redesign based on the HeartFlowAI design system. We've rebuilt the interface around a focused reading experience, simplified navigation, native macOS visual effects, and a more consistent hierarchy across the application. The updated design uses deep surfaces, subtle transparency, and the violet and magenta accents of the HeartFlowAI identity. Our aim isn't simply to make the application look different. It's to create an environment that remains intuitive as more advanced agent capabilities become available. 6. Built for a broader AI ecosystem Aurora is one part of our larger effort to develop accessible AI infrastructure and applications. Alongside HFAI Cloud, we're working toward an ecosystem where users can access different models, interact with specialized companions, and use AI capabilities through a more unified experience. We believe the next generation of AI software will be defined not only by the intelligence of individual models, but by how effectively those models can be integrated into the tools and environments people already use. HeartFlowAI WorkSpace 0.5.0 is another step in that direction. 7. Available on GitHub The updated HeartFlowAI Harness has been committed and pushed to our public repository. We're continuing development on deeper operating-system integration, agent capabilities, and the infrastructure connecting our applications. Explore the project: github.com/HeartFlowAI/hfai_… Learn more about HeartFlowAI: heartflowai.net We're building toward a future where AI isn't simply something you open, but something you work alongside. HeartFlowAI | $HFAI
7
2
7
284