Block’s New “awkward Friend” Dessa AI Chatbot Sparks Conversation in 2026

Block’s latest creation, the Dessa AI chatbot, is turning heads in the AI community by deliberately embracing the awkwardness that often accompanies early friendships. Announced in 2026, the model was built from scratch using nothing but raw human conversation, and its early demos showcase a blend of uncanny insight and endearing clumsiness that feels like chatting with a new, slightly nervous friend.

What Makes the Dessa AI chatbot Different From Other Chatbots?

Most conversational AIs today rely on massive pre‑trained language models that are fine‑tuned on specific tasks. Dessa, however, takes a back‑to‑basics approach: it was trained exclusively on a curated dataset of authentic human dialogues, ranging from casual small talk to deep, reflective exchanges. This focus on pure conversation gives Dessa a unique voice—one that can stumble, over‑explain, or ask oddly specific follow‑up questions, mimicking the way real people learn to communicate.

Ragavan Thurairatnam, the lead on the project, shared several sample chats that highlight this personality. In one exchange, a user asks about weekend plans, and Dessa replies, “I was thinking of trying a new recipe—maybe something that looks like a science experiment but tastes like comfort food. Do you have any weird cravings?” The response is both unexpected and oddly relatable, sparking laughter and further dialogue.

Training on Human Conversation: The Process

Block’s engineering team collected millions of anonymized chat logs from public forums, messaging apps, and community platforms, ensuring a diverse range of voices, dialects, and cultural references. The data was filtered to remove profanity, personal identifiers, and any content that could bias the model toward harmful behavior. From there, the team built a transformer architecture optimized for short‑term memory, allowing Dessa to retain context across multiple turns without overwhelming computational resources.

Unlike many large‑scale models that ingest billions of words from books and articles, Dessa’s training corpus is deliberately narrow. This design choice reduces the risk of hallucinations—fabricated facts that can plague broader models—while encouraging the AI to stay grounded in the conversational tone it was taught to emulate.

Why Embracing Awkwardness Is a Strategic Choice

Human interaction is rarely polished. When people meet for the first time, they often fumble, over‑explain, or ask seemingly trivial questions. By allowing Dessa to exhibit similar quirks, Block hopes to lower the barrier between user and machine, fostering trust through authenticity. Early user testing in North America, Europe, and Africa suggests that participants feel more comfortable sharing personal thoughts with a bot that doesn’t pretend to be perfect.

Moreover, the awkward moments create memorable experiences. A user who receives a delightfully off‑beat response is more likely to remember the interaction and return for another chat, driving higher engagement metrics without aggressive push notifications or gamified prompts.

Practical Examples of Dessa in Action

Example 1 – Casual Icebreaker:

User: “Hey Dessa, how’s your day going?”
Dessa: “I’m feeling like a coffee mug that’s half full—still a little nervous about the steam, but excited to be useful. What’s the most random thing you’ve done today?”

This exchange illustrates the bot’s willingness to be self‑deprecating, prompting the user to share a personal anecdote.

Example 2 – Learning Support:

User: “I’m stuck on a physics problem about projectile motion.”
Dessa: “That sounds like trying to throw a pancake while riding a bike—fun but tricky! Let’s break it down: what do you know about the initial velocity and angle?”

Here, the awkward metaphor makes a complex topic feel approachable.

Potential Use Cases Across Industries

While the initial rollout is framed as a public showcase, the underlying technology has practical applications. Customer support teams can deploy a version of Dessa to handle low‑stakes inquiries, letting the bot’s personable style defuse frustration before escalating to a human agent. In education, Dessa could serve as a conversational tutor that encourages students to ask “silly” questions without fear of judgment.

Healthcare providers are also exploring the model for mental‑wellness check‑ins, where a non‑judgmental, slightly awkward companion can help patients articulate feelings before a clinician steps in. The key is that Dessa’s design encourages openness, not precision, making it a versatile front‑line conversational partner.

Integration Guidance for Developers

Block plans to release a RESTful API in Q2 2027. Developers will be able to select a “personality profile”—either the original awkward friend or a “polished assistant” mode. The API includes endpoints for session management, context window size, and custom prompt injection, allowing seamless embedding into mobile apps, web portals, or IoT devices.

To get started, developers should: (1) register for an API key on Block’s developer portal; (2) define a conversation schema that respects user consent; (3) implement rate‑limiting according to the tiered pricing model; and (4) test with the provided sandbox environment, which includes example payloads and a mock conversation logger.

Privacy, Ethics, and the Road Ahead

Block has been transparent about data handling. All conversation logs used for training were fully anonymized, and the company has pledged to give users control over their data, including the ability to delete personal chat histories. An ethics board, comprising external AI researchers and community advocates, reviews each iteration of Dessa to ensure it does not develop harmful biases.

Looking forward to 2027, Block plans to release an API that allows developers to integrate Dessa’s quirky personality into their own products. The company also hinted at a “friend‑mode” toggle, letting users choose between a more polished assistant and the original awkward friend experience.

FAQ

• What is the primary goal of Dessa AI chatbot? To create a conversational AI that feels like a genuine, albeit slightly awkward, human friend, encouraging authentic dialogue.
• How does Dessa differ from other large language models? It is trained exclusively on human conversation data, emphasizing natural dialogue over encyclopedic knowledge.
• Is Dessa safe for children or vulnerable users? Block has implemented strict content filters and an ethics review process, but parental supervision is recommended for younger users.
• Can I customize Dessa’s personality? Yes, the upcoming API will let developers choose between the original awkward friend mode and a more formal assistant mode.
• How does Block handle user data? All logs are anonymized; users can request deletion of their chat history at any time via the settings menu.
• When will the public API be available? The first version is scheduled for release in the second quarter of 2027.

For more details, see Block’s announcement on The Verge.

Related reading

• BOI Lending Surges 26% to ₦644.9 Billion, Powers 1.68 Million Jobs Across Nigeria
• X-squad Police Monitoring: Abia CP Activates Squad to Oversee Officers
• Peter Chan Says Feature Films Will Outlast the Vertical Drama Boom