
design guy @usebland
San Francisco, CA
Joined October 2011
- Tweets12.1K
- Following470
- Followers680
- Likes44.2K
i've been building my own design benchmarking tool to compare models and efforts, super useful for seeing which models have the best design sense with no skills or customization.
6.2 Astra (Max effort) is the best so far. much better than the same model at Ultra effort, and a lot better than any Anthropic model at any effort level.
to control variance a reasonable amount, i put together a handful of prompts, a local version of our codebase, and an core chat that administers the tests to subagents with a vibed UI to compare the results easier.
having a baseline of our own code, prompts, and a blank harness really helps me compare. the release notes for new models are always so impressive but lack the grounding necessary for me to know if I'm actually using the best model for the project.
1
51
the future is here and it sounds just like a human
Introducing Tofu-1, Bland’s new speech-to-speech model.
Every sound you hear in this video, the caller, the agent, and the background sounds, was generated by Tofu-1.
High-stakes calls don’t happen in quiet rooms. They happen in storms, on the side of highways, with kids crying in the background. Our model was trained on the callers that enterprises talk to every day, so your agents can handle them too.
Join the research preview: bland.ai/speech-to-speech?ut…
3
91
Agamemhuang
Breaking: Nvidia’s board is preparing to oust CEO Jensen Huang after he had two extramarital affairs, including one with a suspected Chinese spy. on.wsj.com/4jcpr01
1
107
raj the 🐐! king of building delightful features out of super complex tech
Today, Bland is sharing our texting agent for business offering. We instantly create an agent across voice and text for your business. Branded, personal, and always on. Check it out: bland.ai/blog/bland-imessage…
117
your URL → your agent, freaky fast 💨
iMessage agents for your business in 30s.
Send a link and Bland learns your business, plugs into your APIs, and creates a branded agent for you automatically.
The agent can do things like Tapbacks (❤️,👍,❗), send embedded components, and even accept payments within a chat.
It's live now. Text +1 (415) 481-7537.
2
129
my baby is live!
so so so excited to launch our redesigned dashboard— many months of hard work and dedication from the entire team to make this a reality 🅱️📈
It’s officially time to kill the phone tree.
From day 1, we bet that our models would one day be good enough to lead the conversation. Today, they can.
So we’ve reimagined how you build agents, putting the model at the center. Pathways no longer orchestrate the whole conversation. They’re one resource your agent draws on, alongside tools and knowledge.
We didn’t stop there:
• Build agents straight from your favorite coding agent
• Deploy across phone, messaging, web, and API
• Test changes with A/B experimentation and reimagined evals
The result is agents that are more flexible, easier to maintain, and built to improve with every model upgrade.
The new dashboard is live today.
Customers have until November 15th to migrate their agents. Self-serve customers can do so using one of our migration tools. Enterprise customers will be migrated by your implementation partner.
Learn more: bland.ai/blog/introducing-da…
4
8
457
they said we gotta go to the dentist 😔 we've got too much plaques 💎🥶
The certification wall is getting a little crowded.
FedRAMP 20x Class A joins SOC 2 Type 1 & 2, ISO 27001, PCI DSS, HIPAA, and more.
Bland’s voice AI infrastructure is now available to federal agencies.
bland.ai/government
1
7
403
what’s the over/under line on when waymo starts selling their camera data to police departments
1
1
190









