BlackBird Beta 1.0 is currently available for macOS. It supports macOS 14 or later, and it’s tested on modern macOS versions. You can try older version, but we cannot guarantee it will work. We recommend running BlackBird on an Apple Silicon Mac (M1-M4 series) or a Mac with a capable GPU for the best performance, though it can fall back to CPU-only mode if needed. In fact, BlackBird’s LaserTune technology is optimized to work even on consumer hardware with minimal GPU memory​, so you don’t need a high-end machine to get started. That said, having at least 8 GB of RAM (16 GB+ preferred) will help for larger AI models. Ensure you have a few gigabytes of free disk space as well, since AI models and any fine-tuned data will be stored locally. (Note: A Windows version is in the works – Windows support is coming soon – but as of now, you’ll need a Mac to use BlackBird.)

All AI processing – from answering questions to fine-tuning models – happens on your device, so your data stays private​.

Currently, BlackBird is Mac-only. The Beta 1.0 is for macOS (with Windows support “coming soon” as the team has announced). The developers are actively working on a Windows version, and it’s expected in the near future. (LaserTune is built to work on both Mac and Windows environments​, so the Windows release is on the roadmap.) As for Linux or other platforms: there’s no official word yet. The focus is on Mac and Windows for now, given those cover the majority of users. If you’re interested in other platforms, keep an eye on Decompute’s announcements or join their community to express interest. For iPhone/iPad (iOS/iPadOS) – BlackBird is a desktop-class application and currently there’s no iOS version. The on-device AI approach could theoretically work on an iPad Pro, for example, but that’s not in Beta at this time. So, for now, plan to use BlackBird on a Mac, or on Windows when that version is released.

BlackBird is an AI agent platform that runs entirely on your Mac. It allows you to build custom AI “agents” (think of them as specialized AI assistants) for different domains like law, tech, finance, etc., without relying on cloud servers​.

BlackBird’s core technology (called LaserTune) even lets you train AI models on your own data locally, tailoring them to your needs without any cloud computing.​

BlackBird supports macOS 14 and later. It works on both Apple Silicon Macs (M1-M4 MacBook Air/Pro, Mac Mini, iMac, etc). Apple Silicon Macs are preferred for better speed (they have Apple’s Neural Engine and excellent CPU/MPS performance for AI), but Intel Macs with sufficient RAM or an eGPU can run BlackBird too. If you’re on an older macOS (pre-14.0), you’ll need to upgrade your OS to install BlackBird. For best results, use macOS 14 or 15. Also, keep your macOS updated if possible, as BlackBird may take advantage of the latest performance improvements in macOS (especially for machine learning libraries).

You can download BlackBird from the official Decompute website (via the BlackBird Beta program link or the public download page provided). Once you have the installer (typically a .dmg file), open it and drag the BlackBird app into your Applications folder. On first launch, BlackBird will ask for permission to access your location – make sure to allow this (it’s needed for region compliance; see Location Policy below). If macOS shows a warning (e.g. “unidentified developer”), you can bypass this by right-clicking the app icon, selecting Open, and confirming you want to open it. After launching, BlackBird will guide you through an initial setup or onboarding process. This might include selecting or downloading an AI model for the first time and walking you through creating your first agent. The onboarding is designed to be user-friendly, so you can create a custom AI agent in just a few steps​.

With BlackBird, you can quickly set up an AI agent to, for example, help summarize legal documents, answer coding questions, analyze financial data, transcribe meetings, or assist with research – all in a few clicks and without needing extensive AI expertise​

BlackBird is designed to run offline. Once it’s installed and you have your AI model set up, you do not need internet for the core features – all inference (AI answering questions) and training happens on your machine​. This means you can use your custom AI agents anywhere, even with no Wi-Fi. However, there are a couple of cases where internet may be used: Initial setup/model download: If BlackBird needs to download an open-source model for you (e.g. pulling a model from an online repository), you’ll obviously need internet for that download. You can also load a model file manually if you already have one, to stay offline. Location verification: On first launch (and occasionally as needed), BlackBird checks your location to ensure you’re in a supported region. This might use an internet-assisted service to determine location (since most Macs don’t have GPS). It’s a quick check and minimal data usage. Aside from those, all your queries, data, and interactions with agents do not require internet – you have an unlimited, offline AI that won’t stop working even if your connection does​.

All processing is local, which is crucial for finance: sensitive financial information (like private ledgers or confidential analyses) stays on your machine​. The Finance agent effectively gives you a private financial analyst that can work with your data anytime. As with all agents, you’ll want to verify critical outputs, especially numerical ones, but it’s a powerful tool to quickly get insights from financial data that would otherwise require a lot of manual spreadsheet work.

Yes! BlackBird includes a Voice Memos & Meetings agent specifically for this purpose. This agent can take an audio recording – for example, the recording of a meeting, a lecture, or even a voice memo you dictate – and transcribe it into text, then do useful things with that text. The workflow is typically: you either record audio directly in BlackBird (if the app provides a recording interface) or import an audio file. The agent will use an on-device speech-to-text model (such as Whisper or a similar AI) to convert the speech to text. This transcription step is done locally on your Mac, so even the audio never leaves your device. Once the meeting is transcribed, the AI can generate a summary of the conversation, list out action items or decisions, and allow you to query the transcript. For example, after a long meeting, you could ask, “What topics did we discuss regarding Project X and what were the conclusions?” and the agent will reference the transcript to answer. This is immensely helpful for reviewing meetings without having to re-listen to them or read raw transcripts. You can also use the voice agent for personal voice notes – imagine you ramble some ideas into a voice memo, and BlackBird’s agent later gives you a neat written summary or even turns it into a structured outline. All of this happens with privacy in mind (no cloud transcription service is involved). It’s like having a smart meeting assistant that takes notes and summarizes for you, right on your Mac. Just keep in mind that the initial transcription might take a bit of time depending on the length of the audio and your hardware, but once transcribed, interacting with the content is fast and all local.

For example, you could provide it with a company’s quarterly financial report (or even feed in structured data like a CSV of financial metrics) and ask, “Summarize the company’s Q1 performance,” or “What are the trends in revenue over the last 5 years?” Because BlackBird can handle text + tabular data in fine-tuning​, the Finance agent can be trained to understand numbers in context – something very useful for finance professionals. You might fine-tune it on historical financial data of your business, enabling it to answer questions like “How does this year’s Q3 compare to last year’s Q3?” without you manually crunching the numbers. The agent can also assist in more general finance tasks like explaining financial concepts, calculating projections, or analyzing investment portfolios if you give it the relevant data.

The General Q&A agent is a catch-all, versatile AI assistant for everyday questions and tasks. If you don’t need a specialized domain agent, you can use the general one much like you’d use ChatGPT or another assistant – with the big difference that BlackBird’s general agent runs locally (no API calls or usage limits). You can ask it a wide range of questions: from trivia (“What is the capital of Brazil?”) to advice (“Give me some tips for managing my time better.”), or even creative prompts (“Help me draft a friendly email to my team”). The general agent uses the base AI model’s broad knowledge. BlackBird allows unlimited queries to this agent since it doesn’t rely on any external service​– you won’t be stopped by rate limits or subscriptions. Do note that if your base model is an open-source LLM without updated internet knowledge, it may not have information on very recent events or very niche topics. But for a lot of use cases (brainstorming, writing assistance, general knowledge Q&A, etc.), it’s extremely handy. Essentially, the General Q&A agent is your AI companion for anything, ready to help even when offline. If you do require up-to-date info on something, you’d currently need to provide that info to BlackBird (since it won’t fetch from the web), but otherwise it’s quite capable on its own learned knowledge.

The Research agent in BlackBird is intended to help scholars, students, or analysts in digesting and querying research materials. Suppose you have a collection of research papers or lengthy reports – you can feed those into the Research agent (either by fine-tuning the model on a corpus of documents, or possibly by using the agent to analyze one document at a time). The agent will then be able to answer questions about the content, summarize sections, and help draw connections between concepts. For example, you might give it a scientific paper and ask, “What are the key findings of this study?” or “How does this experiment compare to previous work in the field?” The advantage of BlackBird here is that you can customize the agent to your specific research area. By fine-tuning the model on several papers from your field, the agent effectively becomes knowledgeable about that domain’s terminology and background. It will capture patterns and important details from those papers​, making it more accurate when you ask niche questions. This is more powerful than a generic AI which might not know about, say, a specialized subfield or might hallucinate; your fine-tuned Research agent will base its answers on the actual material you provided. And, of course, any unpublished research data or confidential reports you use stay on your machine (no risk of leaking your in-progress thesis or a competitive analysis)​. You can also use the Research agent in a more ad-hoc way: drop in a PDF, get a summary, ask a few questions, then drop in another, etc., if you don’t want to fine-tune permanently. It’s all about making it easier to work through large amounts of information. Just remember that while it can summarize and analyze, it’s always good to double-check important facts or numbers against the source documents.

The Technical domain agent is like your personal on-device coding assistant or IT expert. This agent can help with a range of technical tasks: answer programming questions, explain code snippets, help troubleshoot errors, or summarize technical documentation. For instance, you could paste a block of code and ask, “What does this code do?” or “Find any bugs in this function,” and the agent will analyze it and respond. If you’re working with documentation (say, an API reference or a technical manual), you could fine-tune or feed that into the agent so it can answer specific questions about it. BlackBird is particularly useful here because it supports fine-tuning on structured technical data and documentation – it can learn from your specific tech stack or product docs, not just generic programming knowledge​. This means the agent could become specialized in, for example, your internal software project. Developers might use the Technical agent for help with syntax, suggestions for how to implement a feature, or even as a pair-programmer for brainstorming solutions.

Since everything runs locally, you can paste proprietary code or logs into it without worrying about leaking that sensitive information. It is worth noting that the quality of answers will depend on the model and any training you provide; a larger model or a model fine-tuned on code (like a Code LLM) will perform better for complex coding queries. BlackBird lets you choose those models, so you have flexibility to optimize your Technical agent.

Creating an agent in BlackBird is straightforward and no coding is required. When you open BlackBird, you’ll have an option to create or add a new agent. You’ll typically choose a domain or type for your agent (for example: Legal, Technical, Finance, etc.), give your agent a name, and select which base AI model it should use. BlackBird will then set up the agent for you – this might include loading the model into memory and preparing any domain-specific settings. If you have specific data you want the agent to learn from (say, a set of documents or notes), you can provide that to fine-tune the agent (more on that in Model Hosting & Training). Otherwise, the agent will start with the general knowledge of the base model or any built-in domain knowledge provided. You can create multiple agents for different needs – for example, one agent tuned with your legal contracts, another with your programming documentation, etc. There’s no hard limit on the number of agents; it’s constrained only by your storage and memory (since each agent might load its own model or fine-tuned variation). Once created, an agent opens in a chat-like interface: you can ask it questions or give it commands related to its domain. BlackBird’s interface lets you switch between agents easily, so you can have your “Legal AI” and “Tech AI” both available and swap as needed.

Blackbird is not a lawyer – you should always have a human legal professional review any critical outputs. But for day-to-day productivity in a law practice or legal research, the Legal agent can save a lot of time by quickly answering questions like “What does section X of this contract imply?” or “List the main obligations of Party A in this agreement,” all in natural language

Because BlackBird runs entirely locally, it’s suitable for confidential legal work – your sensitive case data never leaves your computer​. Keep in mind, while the Legal agent can be very helpful (e.g. speeding up document review or giving you a first draft summary),

A Legal domain agent in BlackBird is designed to assist with legal documents and questions. For example, you could upload or paste the text of a contract, case law, or a legal brief, and then ask the agent to summarize key points, identify specific clauses, or answer questions about the document. The agent can also help draft legal-style text (like memos or simple contracts) based on prompts you give. Under the hood, if you fine-tune it with your firm’s data or legal datasets, it essentially becomes familiar with that material – capturing patterns in legal language, terminology, and structure. This can be more powerful than just keyword search because the AI actually learns the context​.

BlackBird’s Finance domain agent is geared towards financial data and analysis. This agent can help interpret financial reports, statements, spreadsheets, or economic data.

Upon creating an agent, BlackBird might offer you a default model to start with (something reasonably sized for your Mac). You can swap the model or load a new one via the settings or a model manager in the app. Keep in mind hardware limitations: larger models (like 30B or 70B parameter models) might not run well (or at all) on a typical laptop due to RAM/VRAM limits. As of the current version, models up to around 13B parameters are the sweet spot for many Mac users. The developers have mentioned that support for 15B+ parameter models on laptops is coming in an update (and some users with powerful desktops might already experiment with larger ones).

LaserTune is Decompute’s proprietary fine-tuning technology that is built into BlackBird. It’s what enables you to train or personalize AI models locally on your machine. Fine-tuning means taking a base model and further training it on specific data so it better understands that data or task. Traditionally, fine-tuning large models required powerful servers or cloud GPUs, but LaserTune is optimized for minimal memory usage and efficiency on consumer hardware​. In practical terms, LaserTune allows BlackBird to update an AI model’s knowledge using your data (documents, transcripts, etc.) without needing to send anything to a cloud service. To fine-tune a model in BlackBird using LaserTune, you would typically do the following: go to your agent’s training or data section, provide the data you want to train on (for example, you might select a folder of text files, PDFs, or paste in training examples), and then initiate the fine-tuning process. BlackBird handles the rest – you don’t have to write any code. The app will process your data and adjust the AI model’s weights accordingly (this might take a little time depending on how much data and the model size, but LaserTune makes it as fast as possible). Once done, your agent’s model is now “tuned” to your data. For example, if you gave it a bunch of legal case documents, the agent will now be much more knowledgeable about those cases and use that in its answers. If you fine-tune your company’s internal wiki and support tickets, the agent will learn your company’s lingo and info to answer employees’ questions. This tailoring is done completely offline – your data never leaves your device during training​. One of the key benefits of LaserTune fine-tuning is that it can capture subtle patterns in your data better than just retrieving documents on the fly​. It essentially imprints the knowledge into the model. Another benefit is privacy and compliance: for sensitive domains like healthcare, finance, legal, etc., you can train the model on real sensitive data with no compliance breach because nothing goes out to a third party. LaserTune was noted to have an extremely low memory footprint, meaning even if you don’t have a top-of-the-line GPU, you can still fine-tune a decent-sized model – it’ll use your CPU or any available GPU efficiently​. The process might run slower on pure CPU but it’s doable; on an M1/M2 Mac, it can leverage the Neural Engine/GPU. In summary, LaserTune is what empowers you to personalize your AI agents. To use it, just feed BlackBird your data and hit train – in a few minutes (or longer for big models/datasets), you’ll have a custom model. BlackBird’s interface will likely show you progress and notify you when training is complete. After fine-tuning, you can immediately chat with the agent and it will respond using the new knowledge. And if you ever want to update the training (say you have new data), you can run LaserTune again or do an incremental training. This ability to iterate quickly is why we call BlackBird “agentive” – you, the user, are in control of teaching the AI new tricks as needed.

All models you use in BlackBird, including any you fine-tune, are stored locally on your Mac (typically in BlackBird’s application support directory). This means that when you download or add a model, it’s saved on your disk. Fine-tuned models or any artifacts from training are also saved locally (usually as separate files or checkpoints, so that your original base model remains intact and you have a new tuned version). BlackBird may provide a UI to manage these – for example, a list of downloaded models with an option to remove ones you don’t need, since models can take a lot of space (a 7B model can be ~4GB, 13B maybe ~8–10GB depending on format). If you need to free up disk space, you can delete unused models. Within the app’s settings or the agent settings, look for a “Models” or “Storage” management section. If one isn’t available yet (say you’re on an early beta without that UI), you can manually find the files on disk. They might be located in ~/Library/Application Support/BlackBird/ or a similar path (the exact folder could be mentioned in the docs or by asking the support community). You could delete model files there, but caution: manually deleting files should only be done if you know what you’re doing, to avoid corrupting an agent setup. Ideally, use the app interface to remove them. BlackBird does not upload these files anywhere, so you are in full control. It also means you might want to back them up if you’ve spent time fine-tuning a model and don’t want to lose that work – copy the fine-tuned model file to a safe place or include it in your regular backups of your Mac. As BlackBird evolves, expect more user-friendly controls over local models (for instance, the team has indicated they are working on supporting larger models and that likely comes with better model management UIs). For now, know that everything is stored locally and can be managed either via the BlackBird interface or via the file system.

The time can vary quite a bit based on a few factors: the size of the model, the amount of data, and your hardware capabilities. Smaller models (e.g. 7B parameters) will train faster than larger ones (13B or more). If you only give a few documents or a small dataset, it could fine-tune in just a few minutes. If you provide a very large corpus, it could take longer (tens of minutes to an hour or more). LaserTune is optimized to be faster and more memory-efficient than typical training, so it’s likely significantly quicker than traditional fine-tuning methods on the same hardware. For example, early tests have shown LaserTune achieving fine-tuning on consumer GPUs that normally would have required a high-end server. On an M1 MacBook Air, you might fine-tune a model on, say, 50 pages of text in under 10 minutes (this is a rough guess – actual performance may vary due to memory available and the actual content of the data). On a powerful M2 Pro or a desktop with a good GPU, you can handle larger models or datasets more swiftly. BlackBird will use any resources available – if you have a GPU it can use (like Apple’s GPU via Metal, or an Nvidia/AMD GPU on Windows when that’s out), it will utilize it; if not, it’ll use a multi-core CPU. The interface might not always show the exact estimated time, but it should let you know when it’s done.

The platform is model-agnostic as long as the model is compatible with the backend BlackBird uses (for example, models in the Hugging Face Transformer format or in a quantized .ggml format for efficiency). BlackBird is built to support a variety of open-source AI models for language tasks. It doesn’t rely on a proprietary model – instead, you can choose from supported LLMs (Large Language Models) that can run locally on your hardware. For instance, BlackBird works well with models like Llama 2 and its variants, and the team has demonstrated using third-party models such as DeepSeek and Qwen within the platform​. In practice, this means you could load a model like Llama-2-7B (7 billion parameters) or Llama-2-13B, or other community models of similar size, and use them as the brains of your agents.

No, you don’t. One of BlackBird’s core design goals is to make advanced AI accessible without requiring users to write code or have machine learning expertise​. Fine-tuning an AI model with LaserTune in BlackBird is a point-and-click process. The app provides a user interface for all the steps: selecting data, starting the training run, and monitoring progress. You won’t be writing Python scripts or using command-line tools; BlackBird handles the technical back-end. This means that a lawyer, a financial analyst, or a researcher – even with no coding background – can train their AI agent by simply providing the relevant documents or data through the app UI. BlackBird essentially automates the ML pipeline under the hood. Of course, if you are more technical, you might be curious about what’s going on. But rest assured, nothing beyond basic computer usage is needed to utilize these features. The simplicity does not mean it’s not powerful: behind the scenes, LaserTune is doing sophisticated optimization to fine-tune the model​, but all you see might be a progress bar saying “Training in progress…”. After it’s done, you can immediately test your newly trained agent in the same interface. So whether or not you have a coding background, you can take full advantage of BlackBird’s training capabilities.

BlackBird allows you to train and fine-tune these models (not just run them), which is a big differentiator – effectively, you’re hosting the model and can modify it on your device. Finally, note that BlackBird focuses on open-source models – you won’t be loading OpenAI’s GPT-4, for example, since that isn’t available to run locally. But the open models are continually improving, and BlackBird lets you take advantage of them fully under your control

Absolutely! Blackbird is design to ensure that you can continue using your computer for other tasks during fine-tuning; The underlying LaserTune will consume resources but it won’t completely lock up your system. And since it’s local, you’re not waiting in a cloud queue or anything – you have full control to start or stop it as needed. In short, fine-tuning times are reasonable for practical use. For most customization tasks (a handful of documents to teach the agent), you’ll be done in minutes. For heavy training (like a novel-length dataset), maybe longer. BlackBird’s ongoing optimizations aim to continuously improve training speed in future updates too.

By default, when you create a new agent, BlackBird will assign a default model to it (which you likely downloaded during onboarding). If you want to use a different model, look for a “Model” option in the agent’s settings. There should be an interface to select or import models. Typically, you might have a list of available models that you’ve downloaded. If the model you want isn’t in the list, BlackBird might provide a way to import it – for example, by specifying the model file path or by connecting to an online repository to download it. If you have a model file (like a *.bin, *.pt, or *.ggml file depending on format) you can often drag-and-drop it into BlackBird or use an “Add Model” dialog. Once added, you can select that model for your agent. The agent may need a moment to load the model into memory. After that, any questions you ask that agent will be answered by the newly loaded model. You can change models to experiment with which one gives the best answers for your use case. For example, a smaller 7B model might be fast but less accurate, whereas a 13B model might give better answers but run a bit slower – BlackBird lets you make that trade-off as you see fit. Remember that if you fine-tuned the previous model (say you trained it on your data), switching to a new base model means you’d likely want to fine-tune that one as well on the data, because the fine-tuning doesn’t carry over between different base models. In summary, loading/changing models is intended to be user-friendly in the UI, reflecting BlackBird’s goal to make AI accessible without deep technical steps.

No – BlackBird is built on a privacy-by-design principle. All your data stays on your device​. When you ask an agent a question or provide it with documents, that information is processed locally by the AI model on your Mac. It is not sent to any cloud server for processing​. This is a key difference from many other AI assistants. For example, if you use a cloud AI service, your queries and possibly your documents would be uploaded to a server (raising potential privacy concerns). With BlackBird, the AI lives on your Mac, so your inputs and the model’s outputs never traverse the internet. Even the fine-tuning (LaserTune) process operates locally, meaning if you train on proprietary data, that data isn’t leaving your machine at all. The only times BlackBird might connect to an external service are for things like downloading a model (if you choose to download one through the app) or checking location as discussed below – but it does not send your content or conversations out. This local-only approach makes BlackBird suitable for sensitive industries like finance, law, or healthcare, where keeping data in-house is mandatory​. In summary, you can trust that using BlackBird is as private as storing files on your own computer – because that’s literally what it’s doing, using your own compute resources for AI.

At present, BlackBird leverages the security of your system. This means if you have FileVault disk encryption enabled on your Mac (which encrypts your whole drive), then all BlackBird data at rest on disk is encrypted through that mechanism. BlackBird’s own files are likely not additionally encrypted with a separate key (since it’s running locally for you), but they reside in your account’s data directories. We strongly recommend enabling FileVault (the built-in macOS full-disk encryption) if you are concerned about a lost or stolen device – this will protect all your files, including BlackBird’s, from unauthorized access.

Because of its on-device architecture and privacy focus, BlackBird is especially suitable for domains that handle sensitive information. In law and finance, confidentiality is paramount – BlackBird ensures that client data, contracts, financial statements, etc., never leave your secure environment​. In healthcare or medicine, if you were to use BlackBird to analyze medical texts or patient data (say you’re a doctor who wants to summarize patient visit notes with an AI assistant), you’d remain compliant with privacy laws since none of that data is being transmitted to a third-party (you’d still need to ensure your device is secure per HIPAA guidelines, but the AI isn’t adding new exposure). BlackBird basically flips the typical AI model: instead of sending data out to an AI, it brings the AI to your data. This decentralized approach to AI is inherently more private and secure​. Additionally, BlackBird does not require you to share data with a vendor for improving the model. Some cloud AI services retain user queries to improve their models (unless you opt out). BlackBird’s model fine-tuning stays on your machine, so there’s no such sharing. Each organization or user can have their own custom-tuned AI without handing that proprietary data to anyone else. This is a big shift in control – companies can adopt AI in sensitive areas without the usual privacy trade-off. On the security front, aside from data privacy, BlackBird is as secure as any well-behaved Mac app. It doesn’t open network ports or create weird user accounts or anything like that. Always download BlackBird from official sources to ensure you have a legitimate copy (as with any software). The team behind BlackBird is likely continually auditing and improving the security since they know their target users care about it. If your IT department asks: BlackBird keeps data local, uses local system encryption, and only does minimal verified network calls (e.g., location check, update check) over secure channels. Therefore, it can be white-listed in environments where cloud AI services are banned due to data control reasons. In short, BlackBird was practically made for sensitive domains where privacy is non-negotiable, giving those users access to AI benefits without the usual risks.

BlackBird does not need to communicate externally for location verification or to fetch update, but it does not send your content (either fine-tuned model or the data). Any external communication would be minimal and sent over encrypted channels (HTTPS). The app also follows standard macOS security practices. It will request permissions (like Location Services) when needed and will appear in your System Settings > Privacy lists, so you have transparency into what it’s using. If you use the voice features, the microphone input is only processed by the app locally (macOS will show a mic indicator when it’s in use, as usual). There’s no hidden recording being sent out. In summary, BlackBird secures your data primarily by keeping it local and relying on the proven security of your operating system. Make sure your user account is password-protected (as you normally would on a Mac) and consider disk encryption. BlackBird doesn’t introduce new risks like cloud storage or multi-tenant databases that could be breached – your data lives with you. And if at any point you want to delete something, you can: remove the agent or data from the app, and it’s gone from your disk (there’s no server copy you need to worry about). This local-first approach is a major advantage of BlackBird for users who have stringent data privacy requirements.

Yes, absolutely. Voice data gets the same protective treatment as any other data you use in BlackBird. When you record a meeting or import an audio file into BlackBird’s Voice Memos & Meetings agent, the audio is processed locally on your Mac to generate a transcription. That means if you recorded a sensitive meeting (say an internal company strategy meeting), the audio is not being uploaded to any cloud speech service – it’s likely being transcribed by an on-device AI model (or the macOS speech system) right on your machine. The resulting text transcript is stored locally as well, and any subsequent analysis (summaries, Q&A) on that transcript is done by the local language model. BlackBird does not send the audio or transcript outside your computer. From a user standpoint, you might see BlackBird ask for Microphone access the first time you use the voice feature (macOS will prompt for permission). Granting that allows the app to capture audio for transcription. The audio stays in the app’s control and in your file system. It might save the recording and transcript in a folder (so you can refer back to it). If you’re curious, you can find those files on your disk (likely under BlackBird’s data directory). They would just be audio files (e.g. WAV/M4A) and text files. Again, these are not shared unless you choose to manually share them. In terms of security, if the content of the meeting is extremely sensitive, treat the output like any confidential file on your computer – because that’s what it is. BlackBird has effectively given you a powerful transcription and analysis tool that operates under your full control. Many cloud-based meeting assistants exist, but they would send your audio to their servers. BlackBird’s approach avoids that risk entirely by keeping the data local. So you can confidently use it for private meetings, interviews, personal journal voice notes, or anything else, knowing those remain private. If you ever want to delete a particular recording or its transcript, you can delete the agent or the specific entry (the app may provide a way to manage past meeting notes). Deleting it removes it from your disk (and there is no backup elsewhere).

All data that BlackBird uses – such as any documents you import, transcripts of your voice memos, chat logs with your agents, and the fine-tuned models – is stored on your local disk (typically under your user profile’s Library folders). BlackBird will not upload this data, and it’s only accessible to someone with access to your computer.

When BlackBird checks your location, it’s not doing anything invasive. Typically, it will request a location from macOS. macOS might use a combination of Wi-Fi, IP address, and other signals to determine an approximate location (city or region level accuracy). BlackBird receives that info just to confirm your region. It does not send that raw data back to Decompute (the company) in a way that would violate your privacy. At most, BlackBird might contact a licensing server with an “OK” or “Not OK” signal depending on your location. But that would be a simple flag – not your GPS coordinates. In fact, BlackBird likely doesn’t even obtain precise coordinates; it only cares about country code. The app also doesn’t continuously track location; it likely checks at launch or at set intervals (maybe it re-checks every few days or if the app suspects you moved far, though on a desktop that’s less likely than on a laptop). From a user perspective, you won’t notice anything beyond the initial permission prompt and maybe a quick check at startup. BlackBird doesn’t display your location anywhere, and it doesn’t log it in any user-visible file. The developers have no interest in your exact whereabouts – they just need to enforce the region lock. All of this means that your location data is handled minimally and securely. It’s a binary use: either you’re in an allowed region or not. If yes, you proceed; if not, you’re blocked. They likely keep the logic client-side to a large extent (to avoid false positives/negatives and to reassure users that they’re not building a location database). So, to recap: BlackBird asking for location is a one-time setup requirement for region compliance. It uses that permission responsibly and sparingly. It does not store or misuse location information. If you’re within the supported areas, you have full access; if you leave them, the app will respectfully decline to operate, without any extra snooping. Once BlackBird expands globally, this step might be removed entirely or it will simply always pass. Until then, it’s a necessary step to ensure the app is only used where it’s permitted.

BlackBird is currently only offered to users in certain regions (specifically in India and the United States). This is due to policy and distribution restrictions set by the developers. To comply with these restrictions, the application needs to verify that a user is located in an allowed region. That’s why, when you first run BlackBird, macOS will prompt: “BlackBird wants to use your location.” By granting this permission, you allow BlackBird to perform a one-time (or periodic) check to ensure you’re in a supported country. This check is used solely for compliance – to make sure the app isn’t being used in a region where it’s not yet authorized. BlackBird does not use your location for any other purpose. It’s not tracking your movements or doing anything in the background beyond that region validation. In practice, the app likely just determines your country (it doesn’t need pinpoint GPS coordinates for this, just a general location lookup via IP or location services) and then proceeds if you’re in (for example) the US or India. If you are outside the allowed regions, the app will inform you that BlackBird isn’t available in your location and it won’t fully activate. This might be disappointing, but it’s part of the controlled rollout.

As of now, BlackBird’s public release is restricted to India and the USA. Users in those countries can download and use the app. If you are physically located outside these regions, BlackBird will not function (even if you managed to download it). This geofence is likely temporary; the team intends to expand availability as they navigate regulatory and support considerations. If you are in the US or India and travel abroad, you might encounter an issue when you try to use BlackBird in a different country. Essentially, the app might perform the location check and see you’re not in an approved region, which could disable its functionality until you return. We understand this can be inconvenient for frequent travelers; hopefully future updates might handle this more gracefully (for example, maybe allowing short trips, etc.), but for now it’s a strict check. If you’re not in India or the US at all, you’ll have to wait until BlackBird launches in your region. Keep an eye on official announcements. The restricted rollout could be due to compliance with export regulations (AI software can be subject to such rules), or simply a desire to focus on two markets first before scaling up. Regarding privacy of location: BlackBird’s check will likely use either Apple’s location service or IP-based geolocation. It isn’t storing your location history – it’s just checking “Allowed region: yes or no?” at startup. You can verify that BlackBird isn’t continually accessing location by observing the location services arrow in your Mac’s menu bar (it should only appear briefly). In System Settings, you can also see when BlackBird last accessed location. It should correspond to app launch times. If you revoke location permission after initially granting it, BlackBird will again prompt or may refuse to run until you allow it. So it’s best to keep it allowed for seamless use. In summary, BlackBird’s location requirement is purely about regional availability enforcement. It ensures the app abides by its distribution policy (only US/India users). It’s a small trade-off – you have to let the app confirm your country – in exchange for using this powerful tool. The developers are aware that everyone else is eager to try it, and expansion to more countries will likely come after the beta phase or once they ensure compliance with various local laws.

If the app seems to hang or not proceed due to location, here are a few tips: Make sure you are allowed location permission. Go to System Settings > Privacy & Security > Location Services and check that BlackBird is listed and the toggle is ON. If not, enable it and restart the app. Ensure you have an internet connection at least for the location check. The app may need to query location service (which uses the internet). Once the check is done, you can go offline for normal usage. If you are in the US or India and still get a message saying you’re not in a supported region, it could be a geolocation glitch. This sometimes happens if your internet is via a VPN or your IP address is misidentified as being elsewhere. Try disabling VPNs or anything that might mask your location. You can also try connecting to a different network temporarily. If it still fails, contact support – it’s possible there’s an issue on their end or an update needed to the region list. (In early beta, for example, some users in certain US states might have hit an edge case – these things can happen and the devs would want to know.) Quick workaround: Occasionally, simply restarting the app (and your Mac) can resolve weird hiccups with location services. If none of that works and you’re definitely in an allowed country, drop a note to the BlackBird team so they can investigate. For now, the location step is crucial; unfortunately the app won’t function until it passes this check.

Errors can happen, especially in a beta. Common issues and remedies: Model failed to load: Ensure the model file isn’t corrupted. If you downloaded it through BlackBird and it failed, try downloading again (maybe the download was incomplete). If you manually added a model, double-check it’s in a supported format. BlackBird might expect GGML format for some models; if you gave it a PyTorch .pt model, it might not know what to do with that unless it has conversion built-in. Check documentation for what model formats are supported and use those. Out of memory error: If you try to load a model that is too large for your system (e.g., a 30B model on a 16GB RAM Mac), you might get an out-of-memory or allocation error. The solution is to use a smaller model or a quantized version of the model (quantized models have smaller size but slightly reduced accuracy – many open models offer quantized weights). BlackBird may eventually handle this more gracefully, but for now, stick within model size limits. Fine-tuning error: If LaserTune stops with an error, it could be due to input formatting. Perhaps one of your training files is in a format that wasn’t expected (like a PDF that couldn’t be read, or some binary data). Try using plain text data for training (or ensure the data is properly parsed before training). The log or error message might indicate which file or line caused an issue. If you identify the problematic data, fix or remove it and try again. Another cause might be running out of memory during training – if your dataset is huge, try with a smaller subset to see if it succeeds. Compatibility error: If you updated BlackBird but still have old model files or agents, there could be a version mismatch. In such cases, removing and re-adding the model might help. Also, the BlackBird community (Discord) might have notes on specific versions (e.g., “Beta 1.0 requires re-importing custom models”). UI glitch or freeze: If the app UI froze during an operation, it’s possible the process is still running in the background. You might need to force-quit the app and restart. The good news is your data (models, etc.) is saved on disk, so a restart shouldn’t lose anything except the immediate session state. If you encounter a cryptic error and basic troubleshooting doesn’t resolve it, your best bet is to reach out to the BlackBird support channels (see next question). Since it’s a beta, user feedback is valuable to them for fixing such issues. Provide as much detail as you can: what you were doing, what the error said, what model/data was involved. They’ll likely address it in an update.

If you’ve asked your agent a question and it’s not responding: Give it a bit of time. Depending on the model size and your hardware, some responses will take several seconds or more (especially if it’s generating a long answer). Larger models (3B+) on a laptop can be somewhat slow per token. If it’s just a delay, you might see a loading spinner – that’s normal, just be patient. Check if the model is loaded. If the agent was just created or the app just opened, it might still be loading the AI model into memory (which can take some time for big models). The UI might indicate this (e.g.,”Firing up the engines” or “Loading model…”). Wait until it’s fully ready. Ensure you haven’t run out of system resources. If your Mac is low on RAM and swapping a lot (check Activity Monitor if comfortable), the responses could be extremely slow or freeze. Try closing other apps or stopping other heavy processes. Try a simpler query. It’s possible the agent got stuck on a very complex request. If you asked it to produce a loner response or do a very difficult reasoning based or analytical task, it might be churning. See if it responds to a basic “Hello” or a simple question. If it does, then the long query was just heavy – consider breaking big tasks into smaller questions. If the agent truly seems unresponsive (no output at all after a long time), you can reset that agent by clicking on the “New Chat” button on the bottom left. In the worst case, quit and restart BlackBird to clear whatever was stuck.

Running AI models locally is resource-intensive. If you notice BlackBird consuming a large amount of RAM or CPU: This is somewhat expected when a model is loaded. A 7B model can use a few GB of RAM, a 13B model can use double that or more (depending on quantization). If your system has 16GB RAM or less, this will feel significant. We do automatic memory allocation and clean ups, so that it does not impact your experience with BlackBird and other applications. The best practice is to run fewer applications at a time if you’re constrained. You can quit BlackBird when not in use to free memory. Usually, when not using any agent we take around 500 MBs of your memory. On Apple Silicon, the unified memory architecture means if you push the limits, the system will start swapping to disk which slows everything. Monitor memory pressure (Activity Monitor’s Memory tab shows this). CPU usage: If you don’t have a supported GPU, BlackBird will use CPU cores, which can max them out during processing. This leads to heat and fan usage. On a MacBook, you might feel it. To mitigate, ensure good ventilation, and know that once the agent finishes a response, usage will drop. If you’re in for a long session, you could plug in your Mac (some systems throttle on battery to avoid heat). Some models can use Apple’s Neural Engine or GPU via Core ML/MPS. Check BlackBird settings – there might be an option to enable GPU acceleration. If available, that could offload work from CPU and be more efficient. If you absolutely need to cut down usage, use shorter queries or limit the length of responses (some agent settings allow you to set a max tokens for response). Shorter tasks consume less compute. Finally, make sure you’re using the latest version of BlackBird. Performance optimizations are likely to come in updates. The team is very focused on making on-device AI efficient, and LaserTune itself is about efficiency​. – these benefits might show up in runtime performance too. In summary, some resource use is normal, but you have knobs to turn (model size, acceleration, usage patterns) to fit your comfort. Don’t be alarmed by high CPU during a generation – that’s just the AI brain at work! It should subside when idle.

The quality of answers depends on the underlying model and how well it’s tuned to your domain. If you’re finding the answers aren’t up to par, consider these steps: Fine-tune with more data: One of BlackBird’s strengths is letting you train the model on your domain data. If your legal agent is giving generic or wrong legal info, fine-tune it with a few example documents (contracts, case summaries, etc.). This can ground it in the correct domain knowledge. Fine-tuning helps especially to reduce “hallucinations” about niche content because the model weights adjust to what’s real in your data​. Provide more context in your prompt: Sometimes rephrasing your question or giving context improves answers. For instance, instead of “Explain this document,” you might say “Explain the attached document which is a lease agreement between a landlord and tenant, focusing on the termination clause.” The extra detail guides the AI. Check if the agent is mis-categorized: If you accidentally used the wrong agent or model (e.g., using a general model for a very technical question), you might be expecting expertise it doesn’t have. Make sure to use the appropriate agent. If needed, create a separate agent fine-tuned on the specific topic. Combine retrieval with the model: Although BlackBird focuses on fine-tuning over retrieval, you can manually give it facts. For example, copy-paste a relevant excerpt into your question like, “Given the following info [paste a paragraph], answer X.” This ensures the model has the needed info in the prompt. It’s a bit manual, but it can help correctness. Remember, open-source models can sometimes be less accurate than something like GPT-4, especially on very open-ended knowledge. They might make mistakes or even fabricate answers. Always double-check critical information. Over time, we expect BlackBird to support improved models as they come out (and the team might fine-tune some models themselves to be more reliable). For now, use the tools at hand to guide the AI to better answers.

If you downloaded BlackBird directly (not from the App Store), you might encounter macOS Gatekeeper warnings since the app is new. If you see a message like “BlackBird can’t be opened because it is from an unidentified developer,” you can override this easily: Go to System Settings > Privacy & Security. Under the Security section, you should see a note about BlackBird having been blocked. There will be an option to “Open Anyway.” Click that, and confirm. Alternatively, you can right-click (or control-click) the BlackBird app icon in Finder and choose Open from the context menu; macOS will then present a dialog asking if you’re sure – say Open. This is a one-time action. After you’ve done it, macOS will remember and you won’t be prompted again for this version of the app. In future updates, the BlackBird app might be properly notarized or signed to avoid this, but during a beta it’s not uncommon. Rest assured, if you got BlackBird from the official source, it’s safe to run despite the warning. The warning is there because Apple by default blocks apps from non-App-Store developers until you give explicit permission.

The BlackBird team and community are very accessible and ready to help: Discord Community: The fastest way to get support is via the BlackBird Discord server. The developers have invited users to join their Discord for Beta access and support​. In Discord, you can ask questions, report bugs, and get tips from other experienced users. The devs themselves are often active there to assist. Email / Contact Form: You can also reach out through the Contact Us form on the Decompute website. This will likely send an email to the team. If you were part of the Beta sign-up, you might have received emails from a Decompute address – you can reply there for support as well. While email may be slower than Discord, it’s a good option for more formal support or if you need to share logs and don’t want to do that in a public channel. Documentation & Community forums: Check if Decompute’s site has an FAQ or Docs section (they might have documentation pages for BlackBird). There may already be answers to common issues there, and they might expand it as questions come in. Community forums or threads (like a subreddit or GitHub issues, if any) could also be a place to search for your issue. Updates/Announcements: Make sure to follow Decompute on social media (Twitter/X, LinkedIn). They often post updates and tips. For example, when new features or fixes come out, they’ll announce them, which can preempt some support needs by informing you that “hey, feature X is now available in version Y,” etc. When contacting support, be ready to provide details like: your system (Mac model, OS version, RAM), the BlackBird version you’re running (find this in the app menu > About BlackBird), and a description of your issue. If it’s a bug, describing steps to reproduce it helps the team squash it faster. The support team is friendly and understanding – remember, BlackBird is a cutting-edge platform, so they expect and welcome user feedback. Don’t hesitate to reach out; sometimes a quick message on Discord can solve your problem in minutes. Also, because BlackBird is evolving, the answer today might change with tomorrow’s update. Checking the community for the latest info is always a good idea if something isn’t working as expected.

As of the beta launch, a comprehensive user guide might not have been fully published yet, but here are some resources: The built-in onboarding in the app covers the basics of creating an agent. Revisit that if possible, or look for a “Help” section in the app’s menu. Decompute’s website and blog posts (like the announcement post) explain many features and use cases of BlackBird. While not step-by-step guides, they give insight into what you can do​. The Discord community often has pinned messages or channels for how-tos. For example, there might be a #getting-started channel, or users may have shared their own mini-guides for tasks like fine-tuning a model or integrating a new dataset. If you prefer video, check if the team or any early users have posted a walkthrough on YouTube. Since BlackBird is quite new, this might be limited, but it’s worth a search. You can always ask on Discord or other forums, “Hey, is there a guide for doing X?” and someone may point you to a document or provide instructions. The UI of BlackBird is intended to be intuitive, so don’t be afraid to explore it. Many elements have tooltips or descriptions. And as BlackBird moves from beta to a more mature product, expect an official Help Center or knowledge base (perhaps styled like Perplexity.ai’s help center, which this FAQ is channeling). This FAQ itself is part of building that knowledge repository. For now, the combination of this FAQ, community help, and a bit of experimentation will help you master BlackBird.

BlackBird can be installed on multiple Macs (if you have access to the installer or if it’s on App Store tied to your Apple ID). There’s no license key enforcement in the beta that ties it to one machine. The data, however, does not sync between machines because it’s local. So if you set up agents on your iMac and also install BlackBird on your MacBook, they’ll each have their own separate set of agents and models. If you wanted to “transfer” an agent from one to another, you’d have to manually copy the model/data files and set it up similarly on the other Mac. There isn’t yet a built-in sync or export/import for agents (though that’s a suggestion the team might take up in the future). If you get a new Mac and want to move BlackBird there, the process would be: install BlackBird on the new Mac, then copy over the contents of the BlackBird data folder from old Mac to new (and ensure it’s placed in the correct corresponding location). This should bring your agents and models over. Alternatively, re-download models and re-run fine-tuning on the new machine if copying is not feasible. For multiple user accounts on the same Mac: Each user would have their own isolated BlackBird environment. This is actually nice for privacy – if two people share a computer under different logins, their BlackBird usage is siloed to their accounts. It also means if you want to start completely fresh without any of your existing agents, you could run it under a different macOS user account as a blank slate (though that’s rarely necessary; easier to just delete/create agents in one account).

BlackBird’s settings panel (accessed via Preferences in the menu bar) lets you customize several aspects of your experience. While the exact options may evolve, here are commonly available settings: Model Settings: Choose default model for new agents, enable/disable GPU acceleration (if available), set precision or performance mode (some apps let you choose between faster vs. more accurate generation). Agent Behavior: There might be options to set how verbose or concise the AI’s answers should be, or a default tone (e.g., formal vs casual) if you prefer. Possibly an option to reset all agents’ conversations or clear history. Data & Privacy: Perhaps a toggle to delete all user data (for a clean start) or to opt out of any usage analytics (though BlackBird doesn’t really send data out, they might still have a basic analytics opt-in/out). Appearance: If BlackBird has a chat interface, maybe you can switch between light/dark theme or change the font size for readability. Notifications: If the app does background tasks (like a long fine-tune), maybe it notifies on completion – a setting might exist to control notifications or sounds. Location permission: While not exactly a preference (it’s in System Settings), if you ever needed to re-trigger the location check or see its status, you go to Mac’s privacy settings. Within BlackBird, they might show a message if location is disabled with a shortcut to fix it. Since BlackBird is akin to a development platform, there might also be some advanced settings: Developer Options: Possibly toggles for logging or debug mode, if you’re working closely with support to trace an issue. Integration Settings: If BlackBird integrates with any external tools (for example, sending a summary to your email, or connecting to calendar for meeting transcription), settings for those would be here. Currently, not sure if such integrations exist, but they could come (like “Auto-import my Zoom recordings” – just a hypothetical future feature). For most users, the default settings work fine. You’ll likely only tweak things if you have specific needs (like ensuring it uses the GPU if you have one, or limiting resource usage). The UI is meant to be product-focused, not overwhelming, so preferences are usually a short list of meaningful toggles/sliders. Feel free to explore the Preferences menu to see what’s available and adjust BlackBird to your liking.

No, currently BlackBird does not require a user account or login. When you use the app, it runs standalone on your Mac without any sort of cloud account. During the beta sign-up phase, you might have provided your email just to receive the download link or updates, but that’s outside the app. The app itself doesn’t prompt you to log in. This is in line with BlackBird’s offline-first philosophy – you don’t need to be tethered to a cloud identity to use your own AI. So you can think of BlackBird as traditional software: once it’s installed, it’s ready to go for whoever has access to that Mac user account. One thing to note: if multiple people want to use BlackBird on the same Mac, you could either share the same Mac user account (sharing the same agents and data), or each user could install it on their own account. Since there’s no cloud sync, each installation is separate. In the future, the team might introduce an optional account system for things like syncing settings across devices or managing licenses if there’s a paid version, but as of now in the beta/public launch, nothing of that sort exists. It’s refreshingly simple – just open the app and use it. This also means that if you want to back up your BlackBird data, you don’t have an account to rely on – you’d manually back up the files on your Mac (which is a good practice anyway).

Over time, you might accumulate a bunch of agents, downloaded models, and chat history. BlackBird should provide ways to manage these: Deleting an Agent: If you no longer need an agent, you should be able to delete it from within the app (perhaps right-click the agent in a list and choose delete, or a trash icon, etc.). Deleting an agent will likely remove its chat history and any fine-tuned model associated with it. (The base model file may remain if it’s still installed for use by other agents.) Removing a Model: In the model management section (maybe in Preferences or a dedicated UI), you might see all the model files BlackBird has. From there you can remove ones to free up space. For example, if you tried out a large model and don’t want it taking space, you can delete it. If there isn’t a UI, you could remove the file manually as mentioned earlier (though we hope the UI covers it). Clearing Conversations: Some users might want to reset an agent’s memory or clear the chat log (especially if it’s gotten long or if you’re handing your laptop to someone else and don’t want them scrolling through your Q&A). BlackBird might have a “Clear conversation” button per agent. If not, deleting and recreating the agent is a brute-force way. But likely a clear option exists because it’s common in chat UIs. Exporting Data: On the flip side of clearing, maybe you want to export a transcript of a conversation or the summary of a meeting. BlackBird might allow you to copy chat text or save a transcript to a file. This isn’t exactly management, but it’s related to controlling your data. Use that if you need an external copy before clearing. Storage Location: For advanced users – if you wish to move the storage directory (say you want models on an external drive), that might involve symlinking the folder. BlackBird doesn’t likely support changing the directory via UI yet. So by default it’s in your user Library. Just know that if you delete the app, that folder might remain unless you remove it, which is good (to not accidentally lose data). If you truly want to wipe everything BlackBird, you’d delete the app and also its data folder. Cache and Temp Files: BlackBird might cache some things like intermediate results or processed data. If the app has a cache clear option, that can reclaim space and force a fresh re-process if needed. Since BlackBird values privacy, it gives you the controls to manage your data. For example, if you want to discontinue using it, you can delete all your content easily. There’s no mysterious cloud copy lingering. And if you’re just low on disk space, a quick audit of which models you have stored can help – maybe keep only the ones you actively use.

The method depends on how you installed it: If BlackBird was distributed via the Mac App Store (in the future, if not now), then updates will come through the App Store. You’d see BlackBird updates like any other app there. For the current beta distribution (likely via direct download), the app may have an internal updater. Check the menu bar: under BlackBird in the top-left, see if there’s a “Check for Updates” option. Many apps using frameworks like Sparkle will have this. Clicking it will tell you if a new version is available and let you download it. If there’s no obvious update menu, it might be that you’ll receive update notifications via email or Discord. The team might send out a new download link when a big update is ready. Also, keep an eye on the Discord announcements or the blog – they’ll announce new releases and how to get them. It’s important to keep BlackBird updated, especially in beta, because bug fixes and new features (like larger model support, improved performance, etc.) are coming frequently. For instance, support for 15B+ parameter models was mentioned as an upcoming update – you’d want to grab that once it’s out to take advantage of it. If you’re ever unsure about your version, go to About BlackBird in the app menu to see the version number, and compare that with the latest announced version. Updating should be straightforward – likely just downloading the new app and replacing the old one. Your existing agents and data should carry over since they’re stored separately from the app (usually in your Library). But if you want to be extra safe, you can back up your BlackBird data folder before upgrading. Thus far, upgrades have been smooth for users.

Right now, BlackBird is in a free public beta (with region restrictions). The developers have not yet announced pricing details for the final product. Since the question comes up, here’s what we can speculate/expect: During beta, it’s free to use. They want feedback and to build a user base. In the future, possibly BlackBird could adopt a one-time purchase model or a subscription, especially if they offer continuous updates or cloud-enhanced features (although the core is offline, they might have an optional cloud backup or model library service, for instance). They might also keep a generous free tier and have a Pro tier with additional features (maybe access to exclusive larger models or priority support). Given that running locally doesn’t incur server costs for the company, the value proposition of charging would be tied to the software and support itself. Many comparable tools (if any, since BlackBird is quite unique) could charge a license fee similar to how professional software is sold. As of now, since no payment is involved, you can just enjoy the beta. The team will likely communicate any pricing well in advance of ending the free period. They might grandfather beta users into a deal or at least give notice. For the latest info, check their official channels or FAQ on pricing once they update it. If you’re an enterprise or team thinking of adopting BlackBird widely, you might reach out to Decompute for any enterprise licensing discussions. They could be open to that kind of arrangement (sometimes companies will have an enterprise plan separate from individual pricing). But all that is a bit down the road – we’ll have to see. Keep in mind this FAQ answer is speculative and meant to set expectation that free beta won’t last forever, but details will come from the company. For now, there’s no account or payment system integrated into the app, which means you don’t have to worry about any paywalls. Enjoy experimenting with BlackBird and contributing feedback during this beta phase!

It is likely that you are running a MacOS version 14.6 which is unfortunately not compatible with the format.

You can simply click on the “New Chat” button

- Make sure to copy App file from DMG to "Applications Folder" to ensure login is successful