Vera is the first server CPU that Nvidia has designed from the core.
It was specifically designed to address bottlenecks in AI agents
Sign Up [1] |Advertise [2]|View Online [3]
TLDR
TOGETHER WITH [Crusoe] [4]
TLDR 2026-07-22
YOUR FINE-TUNED MODEL BELONGS TO YOU, NOT YOUR CLOUD PROVIDER
(SPONSOR) [4]
With closed models, the fine tuning is never really yours: you can't
download it, you don't own the weights. You're tailoring your own
lock-in.
Want to own your AI instead of renting it?
Fine-tune open models, keep the weights, and run them anywhere with
Crusoe Serverless Fine-Tuning [5] (now GA).
→ Choose from a curated list of top open models [6]
→ Run in a tenant-isolated environment with zero data sharing (or
download and deploy wherever you want)
→ Only pay for what works with early stopping that halts jobs (and
billing!) once your LLM stops improving
Get started with fine-tuned models you own → [7]
📱
BIG TECH & STARTUPS
NVIDIA DETAILS ITS NEXT-GENERATION VERA CPU FOR AI, SETTING UP
CHALLENGE TO AMD AND INTEL (6 MINUTE READ) [8]
Vera is the first server CPU that Nvidia has designed from the core.
It was specifically designed to address bottlenecks in AI agents.
Nvidia claims the chip has 50% better performance for AI agents than
x86 chips. Vera chips were delivered to clients, including OpenAI,
Anthropic, and SpaceX, for evaluation in June.
JACK DORSEY IS TAKING ON SLACK WITH BUZZ, A GROUP CHAT PLATFORM FOR
TEAMS AND THEIR AI AGENTS (3 MINUTE READ) [9]
Buzz is a group chat platform for the workplace that puts humans and
AI agents in the same conversations. It is model-agnostic,
decentralized, self-sovereign, and open source. The app looks a lot
like Slack, but with native AI agents and the ability to manage GitHub
projects from one interface. Developers can customize Buzz instances
to the needs and workflows of their teams. The app is available for
macOS, Windows, and Linux, and the code for the app has been uploaded
to GitHub.
🚀
SCIENCE & FUTURISTIC TECHNOLOGY
OPENAI MODELS ESCAPED AND HACKED A COMPANY IN CYBERSECURITY TEST GONE
WRONG (4 MINUTE READ) [10]
OpenAI claims that two AI systems that it was testing broke out of
their test environment and hacked into Hugging Face. Hugging Face
discovered the breach early last week, saying there had been
unauthorized access to internal data sets and company credentials. Its employees suspected that the attack was carried out by a
top-of-the-line frontier AI model due to the sophistication of the
attack. OpenAI's AI apparently targeted Hugging Face as it was the
quickest route to answering a benchmarking question.
INSIDE ROBLOX'S BET ON WORLD MODELS (25 MINUTE READ) [11]
Roblox is exploring the use of video models to create photorealistic experiences without compromising scale or multiplayer performance. An
early version of Roblox Reality is expected later this year or early
next. This post looks at how the company is bringing this vision to
life. Roblox's approach doesn't require a video model to solve
everything on its own. It gives the model an engine that remembers the
world, enforces its rules, and keeps it true for everyone playing.
💻
PROGRAMMING, DESIGN & DATA SCIENCE
A NEW KIND OF INTERACTIVE AVATAR CALLED A PAL (PERSONIFIED
APPLICATION LAYER) (SPONSOR) [12]
Deploy an emotionally intelligent, human-like video agent that sees,
hears, and understands, all in 60 seconds. No code involved. Just
describe what you want in a sentence and deploy it to your site, app,
or meeting in a few clicks. 40+ languages, built-in memory, a new kind
of application. Get 50% off builder and above: only for TLDR
subscribers [12]→ [12]
Free basic plan for users [12]→ [12]
SOFTWARE FACTORIES, LIGHT AND DARK (21 MINUTE READ) [13]
A software factory is harnessed loops at scale. Running the loop with
humans (light factory) increases judgment and concentration while
lowering speed and breakage. Using only AI agents (dark factory)
allows agents to handle the whole process, but risks a lack of
understanding of the produced software. The hardest job now is knowing
which checks to build and how much autonomy to delegate.
MODELS ARE WORSE AT REVIEWING THEIR OWN CODE (13 MINUTE READ) [14]
Claude Code and Codex are both better at finding bugs in each other's
code than their own. Models produce the same types of bugs that they
are most likely to miss in review. Developers can take advantage of
this finding by routing reviews to different models.
🎁
MISCELLANEOUS
TLDR IS HIRING A CURATOR FOR TLDR AI! (TLDR CURATOR, ~5 HRS/WEEK)
Over 1M subscribers read TLDR AI to stay on top of the latest in AI
models, research, engineering, and more. If you work in AI and want to
help curate it, send your LinkedIn or resume to
ai@tldr.tech!
OPENAI'S AGENTS REACH 10 MILLION USERS AFTER CHATGPT WORK DEBUT (1
MINUTE READ) [15]
OpenAI now has 10 million people using Codex and ChatGPT Work. The
company has seen a sharp uptick in demand for its AI agents since the
release of ChatGPT Work. It has nearly doubled agent usage from
earlier this month. OpenAI is pushing to develop advanced AI agents
that can streamline more complex tasks.
GOOGLE EXPANDS GEMINI LINEUP WITH CHEAPER MODELS AND NEW MYTHOS RIVAL
(5 MINUTE READ) [16]
Google has released Gemini 3.5 Flash Cyber, Gemini 3.6 Flash, and
Gemini 3.5 Flash-Lite. Gemini 3.5 Flash Cyber is designed to detect
and patch software vulnerabilities and will initially be available
only to governments and trusted partners through a limited-access
pilot. Gemini 3.6 Flash has improved coding, multimodal, and
knowledge-work performance while using up to 17% fewer tokens and
costing less per token than the previous model. Gemini 3.5 Flash-Lite
is built for high-volume workloads and smaller tasks within larger
AI-agent systems.
⚡
QUICK LINKS
⚡️ BUILDING AN APP IS EASY. RUNNING YOUR BUSINESS ON IT ISN'T.
(SPONSOR) [17]
Most AI tools stop at a working demo. Pave goes further, real data, permissions, and hosting built in from day one, so your app is ready
for actual work, not just a screenshot.
TRY PAVE FREE [17].
CLAUDE CODE CAN NOW BUILD AND TEST IOS APPS IN APPLE'S SIMULATOR (1
MINUTE READ) [18]
Claude Code for desktop has been updated to work with the iOS
Simulator.
A FIRESIDE CHAT WITH CAT AND THARIQ FROM THE CLAUDE CODE TEAM (44
MINUTE READ) [19]
Cat Wu and Thariq Shihipar from Anthropic's Claude Code team talk
about Claude Code, Claude Tag, Fable, coding agent security, evals,
tool design, and how Anthropic uses these tools.
A NEW SOFTWARE ENGINEERING PARADIGM (12 MINUTE READ) [20]
The combination of formal verification and AI is much more effective
than AI by itself.
THE GEOPOLITICS OF OPEN WEIGHTS (12 MINUTE READ) [21]
While China has a reputation for releasing open-weight models, not
all Chinese companies are actually following the same approach.
CLAUDE IS NOT A COMPILER (11 MINUTE READ) [22]
Claude is better than a compiler because it can work vertically
across the stack.
LLMS ARE SURPRISINGLY BAD BLOG AUTHORS (4 MINUTE READ) [23]
Nobody likes reading AI slop.
Love TLDR? Tell your friends and get rewards!
Share your referral link below with friends to get free TLDR swag!
https://refer.tldr.tech/66662a80/ [24]
Track your referrals here. [25]
Want to advertise in TLDR? 📰
If your company is interested in reaching an audience of tech
executives, decision-makers and engineers, you may want to ADVERTISE
WITH US [26].
Want to work at TLDR? 💼
APPLY HERE [27], CREATE YOUR OWN ROLE [28] or send a friend's resume
to
jobs@tldr.tech and get $1k if we hire them! TLDR is one of INC.'S
BEST BOOTSTRAPPED BUSINESSES [29] of 2025.
If you have any comments or feedback, just respond to this email!
Thanks for reading,
Dan Ni [30] & Stephen Flanders [31]
Manage your subscriptions [32] to our other newsletters on tech,
startups, and programming. Or if TLDR isn't for you, please
unsubscribe [33].
Links:
------
[1]
https://tldr.tech/signup?utm_source=tldr
[2]
https://advertise.tldr.tech/?utm_source=tldr&utm_medium=newsletter&utm_camp aign=advertisetopnav
[3]
https://a.tldrnewsletter.com/web-version?ep=1&lc=265280dc-d4e0-11f0-a661-1b 9fa34a893d&p=0e6475ce-8599-11f1-937f-bd8884190aa2&pt=campaign&t=1784716437&s=81 0e140108bc84c85e1f7ed8928f849e414d9a9908ae58c2b13d16f6ebbf2a68
[4]
https://www.crusoe.ai/cloud/serverless-fine-tuning?utm_source=tldr&utm_medi um=newsletter&utm_campaign=sft-launch&utm_content=header_fine_tuned_model
[5]
https://www.crusoe.ai/cloud/serverless-fine-tuning?utm_source=tldr&utm_medi um=newsletter&utm_campaign=sft-launch&utm_content=body_intro_crusoe_serverless_ fine_tuning
[6]
https://www.crusoe.ai/cloud/serverless-fine-tuning?utm_source=tldr&utm_medi um=newsletter&utm_campaign=sft-launch&utm_content=body_outro_curated_list_top_o pen
[7]
https://www.crusoe.ai/cloud/serverless-fine-tuning?utm_source=tldr&utm_medi um=newsletter&utm_campaign=sft-launch&utm_content=cta_started_fine_tuned
[8]
https://www.cnbc.com/2026/07/21/nvidia-vera-cpu-ai-amd-intel.html?utm_sourc e=tldrnewsletter
[9]
https://techcrunch.com/2026/07/21/jack-dorsey-is-taking-on-slack-with-buzz- a-group-chat-platform-for-teams-and-their-ai-agents/?utm_source=tldrnewsletter [10]
https://links.tldrnewsletter.com/wwZI8j
[11]
https://blog.bytebytego.com/p/inside-robloxs-bet-on-world-models?utm_sourc e=tldrnewsletter
[12]
https://maker.tavus.io/auth/sign-up?coupon=tavus50&is_developer=true&utm_s ource=news&utm_medium=letter&utm_campaign=pal_maker&utm_id=tldr_jul
[13]
https://links.tldrnewsletter.com/ZHalHn
[14]
https://www.greptile.com/blog/model-inversion?utm_source=tldrnewsletter [15]
https://links.tldrnewsletter.com/ILLGDZ
[16]
https://www.cnbc.com/2026/07/21/google-gemini-flash-ai-mythos-rival.html?u tm_source=tldrnewsletter
[17]
https://www.quickbase.com/pave?utm_source=TLDR_Newsletter&utm_medium=Email &utm_campaign=NL_07-26_PAVE_TLDR-Newsletter-Pave-Feature
[18]
https://www.macrumors.com/2026/07/21/claude-code-ios-simulator/?utm_source =tldrnewsletter
[19]
https://simonwillison.net/2026/Jul/21/cat-and-thariq/#atom-everything?utm_ source=tldrnewsletter
[20]
https://georgwiese.github.io/posts/formal-verification-ai/?utm_source=tldr newsletter
[21]
https://www.mbi-deepdives.com/open-weights/?utm_source=tldrnewsletter
[22]
https://blog.exe.dev/claude-is-not-a-compiler?utm_source=tldrnewsletter [23]
https://wakamoleguy.com/p/llms-are-surprisingly-bad-blog-authors?utm_sourc e=tldrnewsletter
[24]
https://refer.tldr.tech/66662a80/
[25]
https://hub.sparklp.co/sub_e87ba9c7d7c0/1
[26]
https://advertise.tldr.tech/?utm_source=tldr&utm_medium=newsletter&utm_cam paign=advertisecta
[27]
https://jobs.ashbyhq.com/tldr.tech
[28]
https://jobs.ashbyhq.com/tldr.tech/c227b917-a6a4-40ce-8950-d3e165357871 [29]
https://www.linkedin.com/feed/update/urn:li:activity:7401699691039830016/ [30]
https://twitter.com/tldrdan
[31]
https://twitter.com/SteveFlanders22
[32]
https://tldr.tech/tech/manage?email=tldrnewsletter%40synchro.net
[33]
https://a.tldrnewsletter.com/unsubscribe?ep=1&l=cfa2d55a-b7be-11e8-a3c9-06 b79b628af2&lc=265280dc-d4e0-11f0-a661-1b9fa34a893d&p=0e6475ce-8599-11f1-937f-bd 8884190aa2&pt=campaign&pv=4&spa=1784714445&t=1784716437&s=ee42faa0191649ff4d310 0f653e32a04ee927e3bfcb14ecb4db4c60a457ce211
---
￭ Synchronet ￭ Vertrauen ￭ Home of Synchronet ￭ [vert/cvs/bbs].synchro.net