Condense launches proxy to cut AI coding agent bills by up to 72%
Condense.chat opened public access to a proxy that compresses coding-agent context before billing. In tests, it cut costs about 66–72% on deep sessions while preserving core prompts and reporting 94.2% answer faithfulness.
🗞 #sponsored @testingcatalog
Condense.chat opened public access to a proxy that compresses coding-agent context before billing. In tests, it cut costs about 66–72% on deep sessions while preserving core prompts and reporting 94.2% answer faithfulness.
🗞 #sponsored @testingcatalog
TestingCatalog AI News
Condense launches proxy to cut AI coding agent bills by up to 72%
Condense.chat's proxy compresses coding-agent context with two in-house models, cutting token bills by up to 72 percent on deep sessions.
❤5🔥1
BYTEDANCE 🔥: Dreamina Seedance 2.5 is set for an upcoming release on CupCut and partner apps, potentially as soon as next week.
Additionally, Seedance 2.5 will be able to generate cinematic videos up to 30 seconds in standard mode, with a beta long-video mode extending clips to 180 seconds.
That would be huge 👀
h/t x@MarsForTech via DevMode
Additionally, Seedance 2.5 will be able to generate cinematic videos up to 30 seconds in standard mode, with a beta long-video mode extending clips to 180 seconds.
That would be huge 👀
h/t x@MarsForTech via DevMode
❤6
ICYMI: xAI debuts Grok Voice Agent Builder for Enterprises
xAI launched Grok Voice Agent Builder, a beta no-code platform for production phone agents. It combines speech, reasoning, tools, guardrails, telephony, and monitoring, with setup in under two minutes from $0.05 per minute.
🗞 #grok @testingcatalog
xAI launched Grok Voice Agent Builder, a beta no-code platform for production phone agents. It combines speech, reasoning, tools, guardrails, telephony, and monitoring, with setup in under two minutes from $0.05 per minute.
🗞 #grok @testingcatalog
TestingCatalog AI News
ICYMI: xAI debuts Grok Voice Agent Builder for Enterprises
xAI launches Voice Agent Builder beta, letting teams create custom phone agents in minutes without coding, with 80+ voices and real-time tools.
❤2👍2🔥1
Mistral releases Leanstral 1.5 open model for proof engineering
Mistral released Leanstral 1.5, an open-source Lean 4 code agent for formal proofs and verified software. Available via API, Vibe, and Hugging Face, it posts strong benchmark results and found new bugs in 57 Rust repositories.
🗞 #mistral @testingcatalog
Mistral released Leanstral 1.5, an open-source Lean 4 code agent for formal proofs and verified software. Available via API, Vibe, and Hugging Face, it posts strong benchmark results and found new bugs in 57 Rust repositories.
🗞 #mistral @testingcatalog
TestingCatalog AI News
Mistral releases Leanstral 1.5 open model for proof engineering
Mistral AI introduces Leanstral 1.5, an open-source code agent for Lean 4 formal proof engineering, now available via Labs API and Hugging Face.
❤4👍3
OpenAI might be preparing GPT-5.6 for next week's release
OpenAI’s GPT-5.6 remains in limited Codex/API preview, with Sol, Terra, and Luna as new tiered models. Codex builds point to a new slider for reasoning depth, while wider release depends on US government review.
🗞 #chatgpt @testingcatalog
OpenAI’s GPT-5.6 remains in limited Codex/API preview, with Sol, Terra, and Luna as new tiered models. Codex builds point to a new slider for reasoning depth, while wider release depends on US government review.
🗞 #chatgpt @testingcatalog
TestingCatalog AI News
OpenAI might be preparing GPT-5.6 for next week's release
OpenAI previews GPT-5.6 in Codex with Sol, Terra, and Luna models, introducing a reasoning slider for speed-depth control and an upcoming release.
❤5
Google tests new Gemini Inbox section for Workspace triage
Recent Gemini builds point to an inbox for Workspace users, with follow-up, done, and review filters. The move suggests Google is shifting Gemini from chatbot to task-focused workspace for email triage and automation.
🗞 #gemini @testingcatalog
Recent Gemini builds point to an inbox for Workspace users, with follow-up, done, and review filters. The move suggests Google is shifting Gemini from chatbot to task-focused workspace for email triage and automation.
🗞 #gemini @testingcatalog
TestingCatalog AI News
Google tests new Gemini Inbox section for Workspace triage
Google is testing a dedicated inbox in the Gemini app for Workspace users, with smart filters to triage, track, and review tasks with AI.
❤3
Meta is rolling out Dynamic Photo feature for Meta Glasses to more users.
This feature allows glasses to capture additional frames so you can select the best one later if needed.
Media templates have also been added to the Media tab recently.
Additionally, a Fitness section is now visible in the settings but doesn’t seem operational yet.
> Log a workout on your connected device to get insights.
This feature allows glasses to capture additional frames so you can select the best one later if needed.
Media templates have also been added to the Media tab recently.
Additionally, a Fitness section is now visible in the settings but doesn’t seem operational yet.
> Log a workout on your connected device to get insights.
❤5👍2
Meta prepares Scheduled tasks for Meta AI users on web
ICYMI: Meta is testing scheduled tasks for Meta AI on the web, letting users run recurring prompts in-app. At this moment, Meta races rivals, amid claims its next model nears GPT-5.5 on internal tests.
🗞 #meta @testingcatalog
ICYMI: Meta is testing scheduled tasks for Meta AI on the web, letting users run recurring prompts in-app. At this moment, Meta races rivals, amid claims its next model nears GPT-5.5 on internal tests.
🗞 #meta @testingcatalog
TestingCatalog AI News
Meta prepares Scheduled tasks for Meta AI users on web
Meta is working on a scheduled tasks feature for Meta AI, letting users set and manage recurring instructions on the web, as spotted in new builds.
❤5👍2👀1
ByteDance set to launch Seedance 2.5 with 3-minute AI video output
ByteDance appears close to launching Dreamina Seedance 2.5, possibly on July 9. Public pages point to 30-second standard generation, extension to 180 seconds, and rollout across Dreamina, CapCut, and partner platforms.
🗞 #bytedance @testingcatalog
ByteDance appears close to launching Dreamina Seedance 2.5, possibly on July 9. Public pages point to 30-second standard generation, extension to 180 seconds, and rollout across Dreamina, CapCut, and partner platforms.
🗞 #bytedance @testingcatalog
TestingCatalog AI News
ByteDance set to launch Seedance 2.5 with 3-minute AI video output
ByteDance’s Dreamina Seedance 2.5 is set to launch in July, bringing up to 3-minute AI video generation to creators.
❤6
xAI is preparing to promote its Voice Agent Builder on Grok, allowing users to build custom agents that can answer their phone calls.
> Let Grok answer your phone
> I’ll help you deploy your first agent in just a few minutes without a single line of code.
> Create an AI that picks up your calls and talks in a natural voice. Tell it what to handle — questions, bookings, messages — and we'll give you a free number so it can start taking real calls in minutes.
That would be interesting 👀
> Let Grok answer your phone
> I’ll help you deploy your first agent in just a few minutes without a single line of code.
> Create an AI that picks up your calls and talks in a natural voice. Tell it what to handle — questions, bookings, messages — and we'll give you a free number so it can start taking real calls in minutes.
That would be interesting 👀
❤6
LongCat-2.0 is now available on Huggingface under the MIT license.
> LongCat-2.0 is a 1.6T MoE 1M-token context model from Meituan.
> Works natively with Claude Code, OpenClaw, and Hermes Agent.
Open testing time 👀
> LongCat-2.0 is a 1.6T MoE 1M-token context model from Meituan.
> Works natively with Claude Code, OpenClaw, and Hermes Agent.
Open testing time 👀
❤6
Tencent released Hy3 (non-preview), a 295B-parameter MoE open model with 21B active and 3.8B MTP layer params.
> This Hy3 release follows the Hy3 Preview launch in late April.
> Hy3 scored 2.67/4 on benchmarks, outperforming GLM-5.1 at 2.51/4. The advantage was clearest in frontend development, CI/CD, and data & storage.
More players are catching up 👀
> This Hy3 release follows the Hy3 Preview launch in late April.
> Hy3 scored 2.67/4 on benchmarks, outperforming GLM-5.1 at 2.51/4. The advantage was clearest in frontend development, CI/CD, and data & storage.
More players are catching up 👀
❤9👍3
OpenClaw for iOS got an upgraded design with a better Liquid Glass support and more.
Community update 🦞
Community update 🦞
❤5👍3🔥3
An upcoming Claude Voice Mode UI got updated again. Model selector is still present at the bottom while other UI elements are scattered around the screen.
> While Claude is speaking, the same Anthropic glow appears as on the desktop.
> Seems like voice samples got updated as well.
Curios to see how non Haiku models will perform there!
Soon?👀
> While Claude is speaking, the same Anthropic glow appears as on the desktop.
> Seems like voice samples got updated as well.
Curios to see how non Haiku models will perform there!
Soon?
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
❤4👍3