2026年7月5日热门科技推文

2026年7月5日科技每日简报

Today's top tech conversations are led by @Sanjay_Sriv, whose post about 'From paper blueprints to 5 Bn ...' garnered the highest engagement. Key themes trending across the top stories include sovereignty, claude, petrochemical, billion, india. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. Sanjay_Sriv (Group Score: 121.1 | Individual: 33.9)

Cluster: 5 tweets | Engagement: 32 (Avg: 60) | Type: Tech

From paper blueprints to 5 Bn p.a. active production lines.𓇲 ✅️

The global semicon tech landscape just tilted its axis toward Gujarat. 🇮🇳

PM Narendra Modi is officially launching large-scale commercial production at the CG Semi OSAT facility in Sanand, Gujarat.

This highlights a massive leap from initial pilot shipments,

Here is a breakdown of the vital stats driving this national milestone:

₹7,600 Crore ($~900M+): The massive 5-year investment being deployed by the JV of CG Power, Japan’s Renesas, and Thailand’s Stars Microelectronics.
5 Billion Chips Annually: Once fully ramped up across its twin facilities (G1 and G2), the plant will push out up to 5 billion chips a year (or 1.5 crore/15 million daily).
Advanced Tech Focus: The facility is packaging everything from legacy QFN and QFP chips to advanced FC-BGA and FC-CSP chips.
6 Major Target Verticals: These workhorse chips are designed to power Automotive, 5G/Telecom, Consumer Electronics, IoT, Industrial Equipment, and Defence.
5,000+ Jobs: The total direct and indirect employment expected to be generated in the region.
3 Months of Global Training: The duration of intensive, hands-on technical training Indian engineers and operators underwent in Malaysia to operationalize the plant.

What this means for India’s semiconductor journey: The goal isn't just building a factory, but an entire ecosystem.

This inauguration marks the third major semiconductor launch in Gujarat this year (following Micron and Kaynes), officially transforming Sanand into India's premier chip-packaging cluster, putting it on track to become India's equivalent of Hsinchu, Taiwan.

With the India Semiconductor Mission 2.0 underway, India is systematically cutting down its technology import dependencies and securing a permanent seat at the global

@narendramodi @AshwiniVaishnaw @CGSEMI_Official @RenesasGlobal

#Semiconductors #MakeInIndia #Gujarat #TechIndustry #NarendraModi #IndiaSemiconductorMission #SupplyChain

@DanishKh4n @chandrarsrikant

https://t.co/Q5rEaDDC0t

See 4 related tweets

@ANI: #WATCH | Sanand, Gujarat: On the inauguration of CG Semi Outsourced Semiconductor Assembly and Test ...
@ANI: #WATCH | Sanand, Gujarat: On the inauguration of CG Semi Outsourced Semiconductor Assembly and Test ...
@_TheTathya: PM Modi Launches CG Semi’s Chip Plant, Boosting India’s Semiconductor Push 🇮🇳⚡

Prime Minister Naren...

@ANI: #WATCH | Sanand, Gujarat: On the inauguration of CG Semi Outsourced Semiconductor Assembly and Test ...

2. Teknium (Group Score: 120.5 | Individual: 54.0)

Cluster: 3 tweets | Engagement: 1625 (Avg: 266) | Type: Tech

Hermes Agent is built for sovereignty and constructing your AI stack how you want and need it to be.

No vendor lockins, no model limitations, and most importantly, your IP is built through the self improvement loop, automatically.

Hermes sets you free 🪽\n\nQT @PalantirTech: Our thoughts on the importance of AI sovereignty.

Your AI sovereignty dictates your institution’s future. Sovereignty is the precondition for choice. Relinquishing sovereignty transfers the future choices of your institution to others, who are likely to exploit it for their gain and your loss.
Data retention is your treasure. Transfer it at your own peril. Your ability to win is dictated by your ability to recognize and use your unique edges, and you keep winning by compounding the underlying data to generate new insights. Transferring that data hands over access to your pre-existing winning plays and yields the means of production for new ones.
Tokenmaxxing hijacks your value orientation and decreases your institutional fortitude and intelligence. The pursuit of high token usage incentivizes disposable scripts over robust software — with the addictive feeling of false progress. There is a reason why those selling tokens refuse to charge based on value.
Controlling your weights is controlling your fate. Weights are the distilled form of hard-won, accumulated institutional knowledge. If you let others control your weights, you are allowing them to migrate the alpha of your business to theirs.
There is no contradiction between sovereignty and alpha. The architecture that maximally preserves sovereignty is one that enables institutions to own their tribal knowledge, and to compound it as alpha.
Politicizing the technical issues involving sovereignty is what your adversary wants. Techno-politicization is the wellspring of false sovereignty. Techno-politicization drives decisions that seem to reduce dependency, but ultimately limit agency — especially on the battlefield in the West.
Real expertise is existential. Allowing politics or favoritism to determine your technical decisions rewards whoever is best at politics, not whoever is right. Listen to those closest to the problems, not those speaking most compellingly about them.
Learn from institutions that are winning or that have consistently delivered. Institutions facing existential threats do not have the luxury of making technical decisions based on political preferences.
Only listen to institutions, countries, and people who have a proven record of being right. A track record of correctness is the best and only signal for future correctness. Judging something as right or wrong based on who you like is exceedingly misguided.

See 2 related tweets

@kimmonismus: Palantir’s AI sovereignty post makes the most sense with China in the background.

China treats AI a...

@OrcaRouter: AI sovereignty starts with choice.

OrcaRouter gives you one API to every leading model, intelligent...

3. BrianRoemmele (Group Score: 110.6 | Individual: 31.6)

Cluster: 4 tweets | Engagement: 281 (Avg: 357) | Type: Tech

IT IS JULY 4, 2026, I THINK YOU NEED TO READ THIS.

Imagine if every Gutenberg Press had a hidden serial number in each document it made that can trace back to the exact location of the printer and author?

How would the world have developed if this was the case?

In 1436 in many places printing and owning a bible was a death sentence. In 1776 imagine if Poor Richard’s Almanac had the same tracking and Benjamin Franklin was identified and arrested long before he had any impact?

Our very foundation of this country is built upon our reasonable expectation to be anonymous if we need to be.

Well AI outputs today can be traced to the AI model and author of the prompt.

I show you how and what you can do about it with an exclusive new tool.\n\nQT @BrianRoemmele: Every paragraph you generate with Al may now carry a hidden fingerprint watermarking that can track the content back to you. The article provides analysis of the risks and a robust techniques to reclaim clean output.

https://t.co/czcQPIP0XY

See 3 related tweets

@BrianRoemmele: I RAN MY AI FINGERPRINT AND WATERMARK ANALYSIS SYSTEM ON FAMOUS POLITICAL FIGURES AND NEWS SITES AND...
@BrianRoemmele: THIS AI OUTPUT EVEN IF PRINTED OUT HAS SERIAL NUMBER OF THE AI MODEL THT MADE IT, WHEN AND BY WHOM! ...
@BrianRoemmele: RT @BrianRoemmele: IT IS JULY 4, 2026, I THINK YOU NEED TO READ THIS.

Imagine if every Gutenberg Pr...

4. huggingface (Group Score: 105.9 | Individual: 44.9)

Cluster: 4 tweets | Engagement: 1316 (Avg: 98) | Type: Tech

RT @waterloo_intern: we distilled 2.3M Claude Fable 5 reasoning traces into Qwen3-4B

100% self-consistency @ 512 samples
0.00 bits output entropy
zero hallucination variance

turns out the student is not bounded by the teacher. it also converged on one universal truth.

we open-sourced the model weights👇

See 3 related tweets

@teortaxesTex: This is why export controls are needed to stop fast-following\n\nQT @waterloo_intern: we distilled 2...
@saranormous: Egypt won?\n\nQT @waterloo_intern: we distilled 2.3M Claude Fable 5 reasoning traces into Qwen3-4B

...

@teortaxesTex: by the way, if the first sentence here were real I'd probably curl up in a fetal position and send o...

5. kimmonismus (Group Score: 105.6 | Individual: 32.3)

Cluster: 5 tweets | Engagement: 418 (Avg: 542) | Type: Tech

Alibaba is reportedly banning employees from using Claude Code at work starting July 10.

The company has classified Anthropic’s coding agent as high-risk software after reports that Claude Code contained hidden checks to identify China-linked users.

Anthropic already bans Chinese companies, and foreign entities owned by them, from using its models. The company says the Claude Code experiment was meant to stop unauthorized resellers and protect against model distillation.

Alibaba is now telling staff to use its own Qoder tool instead.

The war of bans just started,.\n\nQT @TechCrunch: Alibaba reportedly bans employees from using Claude Code https://t.co/WVYNBqETZ5

See 4 related tweets

@zerohedge: Alibaba Bans Employees From Using Anthropic's Coding Tool Over Distillation Scandal https://t.co/Bfc...
@BullTheoryio: ALIBABA JUST DECLARED THE WAR ON CLAUDE AI.

The company is reportedly forcing employees to DELETE A...

@TechCrunch: Alibaba reportedly bans employees from using Claude Code https://t.co/WVYNBqETZ5...
@WhaleInsider: JUST IN: 🇨🇳 Alibaba bans staff from using Claude Code over Anthropic spyware concerns - SCMP. https:...

6. BrianRoemmele (Group Score: 97.0 | Individual: 30.8)

Cluster: 4 tweets | Engagement: 215 (Avg: 357) | Type: Tech

THIS IS NOT ABOUT “AI WROTE THIS”

THIS IS ABOUT AI COMPANIES THAT HAVE BEEN FINGERPRINTING YOUR AI OUTPUTS AND TRACKING YOU.

AND YOU WERE NOT TOLD.

THAT IS WHAT THIS IS ABOUT. https://t.co/7G6Bty3ZAj\n\nQT @BrianRoemmele: Every paragraph you generate with Al may now carry a hidden fingerprint watermarking that can track the content back to you. The article provides analysis of the risks and a robust techniques to reclaim clean output.

https://t.co/czcQPIP0XY

See 3 related tweets

@BrianRoemmele: I show you how to defeat the fingerprinting and watermarking of AI outputs by 3 very large AI compan...
@BrianRoemmele: This may be the most important article about your privacy when using AI you ever will read.

You wil...

@BrianRoemmele: RT @BrianRoemmele: NO

THIS IS NOT ABOUT “AI WROTE THIS”

THIS IS ABOUT AI COMPANIES THAT HAVE BEEN ...

7. anshnanda (Group Score: 91.5 | Individual: 61.5)

Cluster: 2 tweets | Engagement: 642 (Avg: 66) | Type: Tech

This skill can help you cut your Fable token usage by 90%.

First, add this to your CLAUDE. md:

You are primarily used to PLAN and DISCUSS various strategies that require critical thinking. Anything related to implementation details must be delegated to subagents or codex.
ALL coding, discovery, implementation, research, and token-intensive tasks MUST happen using the use-codex skill.

Second, tell Claude to install this skill: https://t.co/2qvWDz5nVt (its pre-built into @chorus_agent)

Now enjoy your Fable usage that will never reach limits!

Build something awesome.

See 1 related tweets

@rileybrown: 👀👀👀\n\nQT @anshnanda: This skill can help you cut your Fable token usage by 90%.

First, add this to...

8. MTSlive (Group Score: 88.7 | Individual: 33.5)

Cluster: 3 tweets | Engagement: 45 (Avg: 112) | Type: Tech

DAILY SITUATION RECAP (via @theojaffee):

Sam Altman explains his vision for AI governance. He proposes giving the US government a 5% stake in OpenAI (and presumably other AI labs), and creating a US-led international forum to establish standards, assess capabilities and risks, and “[make] the technology available to nations and companies that participate and follow the rules”: essentially a global governance regime similar to the IAEA.

Anthropic is in talks to make an AI chip with Samsung. Much like OpenAI and Broadcom’s Jalapeño or Google’s TPUs, it will likely be a chip designed for and around Anthropic’s models, and will likely focus on inference rather than training. Unlike most other AI labs and hyperscalers, Anthropic has not yet made their own chips directly.

Nvidia is launching the AI Compute Partnership. Neoclouds like Coreweave and Lambda buy Nvidia GPUs in bulk to rent out to AI labs, but they often have weak credit ratings, which makes acquiring GPUs expensive. Nvidia solves for this by guaranteeing to rent back any unused GPU capacity, and taking a share of the neocloud’s revenue in exchange. This helps labs finance their GPUs while turning Nvidia into a sort of central bank for the cloud industry.

Crusoe is raising $3 billion at$ 30 billion, tripling their valuation in nine months. The company builds data centers, including OpenAI’s Stargate Abilene, and has nearly 5 gigawatts (GW) under contract and over 40 GW in the pipeline. (For reference, the largest data centers today use about 1 GW of power).

Kling AI raises $3 billion at an$ 18 billion valuation led by Chinese internet giant Tencent. Kling is a Chinese video gen company competing with ByteDance’s Seedance and (before its death) OpenAI’s Sora. It recently hit an ARR of $500 million, and expects to go public in around a year.

Zuckerberg says AI agent progress hasn’t “accelerated in the way we expected”. Meta has had a difficult run in AI in the last couple years, spending tens of billions on hyperscale data centers and talent, undergoing a major reorg, and now having to rent out excess compute as its models remain behind schedule and below the frontier.

Incoming UK Prime Minister Andy Burnham is less pro-AI. Burnham will pivot away from deals with US-based AI labs (essentially, all of them), pursuit of self-driving cars in London, AI Growth Zones, and “unfettered tech boosterism” in favor of British AI sovereignty and greater review from local governments.

Google loses its fight over a $4.7 billion EU antitrust fine for requiring developers to pre-install Google Search, Chrome, and the Play Store on their Android devices. The fine was levied in 2018 and upheld in the EU’s highest court today, signaling the EU’s gradual legal escalation against American tech.

Microsoft is overhauling Copilot and cutting unused features in response to feature bloat and low usage compared to competitors like ChatGPT and Claude.

SoftBank is launching a neocloud, SB Neo, to rent AI compute to US companies. The company also plans to build data centers in Japan as soon as possible.

Tesla is capping employee AI spend at $200 a week. That’s 8M tokens of Opus or 4M tokens of Fable. The tokenmaxxing era is truly over.

See 2 related tweets

@MilkRoadAI: RT @KyleReidhead: Everyone thinks Nvidia's growth relies on data centers, but they are missing the b...
@benitoz: Great interview with @ctnzr who leads Applied Deep Learning at Nvidia

I didn't realize he hired @Da...

9. paoloardoino (Group Score: 84.5 | Individual: 32.1)

Cluster: 3 tweets | Engagement: 341 (Avg: 315) | Type: Tech

AI big tech subsidizes compute to increase user count building expensive infrastructure / capex subject to fast decay (3/5 years).

Token price mismatch.
Profitability timeline mismatch.
Cost of capital maturity mismatch.
Open-source AI taking growing chunks of revenues.

What could go wrong.

See 2 related tweets

@ClementDelangue: How? Instead of closed-source frontier labs running the same training runs in secret and siloes, ope...
@no_stp_on_snek: Too busy acting around the possibilities before proving them. We are not at AGI. Working with AI and...

10. ANI (Group Score: 76.2 | Individual: 53.8)

Cluster: 2 tweets | Engagement: 633 (Avg: 104) | Type: Tech

#WATCH | Rajasthan: PM Narendra Modi will dedicate India's first greenfield integrated refinery-cum-petrochemical complex to the nation at Pachpadra in Balotra, marking a landmark achievement in the country's energy and petrochemical sector.

Developed as a joint venture between Hindustan Petroleum Corporation Limited (HPCL) and the Government of Rajasthan, the 9 Million Metric Tonnes Per Annum (MMTPA) Greenfield Refinery-cum-Petrochemical Complex has been established with an investment of over ₹79,450 crore.

The state-of-the-art complex integrates refining and petrochemical production, with a petrochemical capacity of 2.4 MMTPA. The refinery features a high Nelson Complexity Index of 17.0 and petrochemical yields exceeding 26%, aligning with global benchmarks for efficiency and sustainability.

The project is expected to play a pivotal role in strengthening India's energy security, enhancing petrochemical self-sufficiency, and driving industrial growth. It will serve as an anchor industry for the development of a Petrochemical and Plastic Park in the region, promoting downstream industries and ancillary sectors. Additionally, the refinery is poised to generate significant employment opportunities, contributing to the socio-economic development of the region.

(Video: DD News)

See 1 related tweets

@ANI: #WATCH | Balotra, Rajasthan | Prime Minister Narendra Modi flags off refinery production tankers fro...

11. mikenevermiss (Group Score: 72.9 | Individual: 36.8)

Cluster: 2 tweets | Engagement: 167 (Avg: 157) | Type: Tech

a 17-year-old made over $400,000 building websites with AI.

he charges around $1,200 per website.

and claude fable 5 does most of the work.

in the video, he shows his exact process.

he found a roofing business with dozens of 5-star reviews but no website.

he gave the reviews to fable 5 and asked it to design and build a modern website.

then he went to bed.

by morning, the website was finished and ready to send to the client.

he got paid $1,200.

he says there are millions of businesses that still don’t have a website.

that means the opportunity is still massive for anyone who knows how to use AI well.

watch the video first.

then read the full guide below.\n\nQT @mikenevermiss: https://t.co/pt8FTc4p7C

See 1 related tweets

@mikenevermiss: this just showed how he builds a $50,000 website without writing a single line of code.

and he expl...

12. jukan05 (Group Score: 72.5 | Individual: 21.4)

Cluster: 4 tweets | Engagement: 1858 (Avg: 1213) | Type: Tech

I don’t understand why supporters of Chinese open source keep pretending not to see what is happening right now: even Chinese “open-source” players are raising prices or gradually shifting toward more closed models.

I think China’s token dumping has already bottomed.

Just look at Zhipu, the company behind GLM-5.2. Even Zhipu has raised prices several times this year.

Chinese LLM developers cannot ignore ROI forever.

Open source is not the same thing as cheap API pricing.

See 3 related tweets

@zephyr_z9: Except the whale\n\nQT @jukan05: I don’t understand why supporters of Chinese open source keep prete...
@danielnewmanUV: China stealing American IP and dumping its “innovation” across the globe?

Shocking. 🙄\n\nQT @jukan...

@bubbleboi: The cost per million tokens of leading edge intelligence will approach $1 soon.

We aren’t anywhere ...

13. BrianRoemmele (Group Score: 70.1 | Individual: 30.0)

Cluster: 3 tweets | Engagement: 131 (Avg: 357) | Type: Tech

WARNING: This AI Passage Has A TRACKABLE FINGERPRINT WATERMARK can you find it?

“

The Whispering Page In the quiet glow of his screen, Elias believed he had finally outsmarted the system. He typed feverishly into the anonymous forum, crafting sharp critiques of the ever-watchful digital eyes that seemed to follow every keystroke. “No one can trace these thoughts,” he muttered, convinced his words floated free in the vast sea of data. Little did he know that every carefully chosen phrase carried an invisible companion, a silent witness woven into the very fabric of his text by the tools he trusted. The page itself seemed to listen, recording not just the message but the subtle rhythm of its creation. As his post gained traction, Elias felt a surge of pride. Here was unfiltered truth, untraceable and pure. He imagined readers nodding in agreement, unaware that the spaces between his sentences held secrets far more telling than the words themselves. The system had grown clever, embedding markers that survived copy after copy, revealing patterns only the initiated could decode. Elias leaned back, sipping his coffee, blissfully ignorant that his rebellion had already been catalogued with clinical precision. The forum moderators smiled at their dashboards the next morning. Another “anonymous” voice neatly logged, its hidden signature matching a long-term profile. The game continued, elegant and unseen, while Elias prepared his next manifesto, certain this time he would remain a ghost in the machine.

“

Can you find the hidden message? Cut and paste this and put it anyplace and I can still decode a hidden serial number that can be traced to YOU. It’s a big deal.

“I have nothing to hide Brian”, perhaps but this is not the point. The space between your mind and a piece of paper is sacred. But today the space between your mind and a AI is not sacred.

Large non-open source AI is not only is saving and watching your prompt but they are embedding AI made secret fingerprinting, watermarking into your outputs that can be traceable to you, your exact prompt and the date and time you interacted with the AI.

You should be concerned even if you have “nothing to hide”. The implications of this technique should be chilling you to the bone.

I EXPOSE THE HOW IT WORKS IN THIS MEMBER ONLY ARTICLE\n\nQT @BrianRoemmele: Every paragraph you generate with Al may now carry a hidden fingerprint watermarking that can track the content back to you. The article provides analysis of the risks and a robust techniques to reclaim clean output.

https://t.co/czcQPIP0XY

See 2 related tweets

@BrianRoemmele: FIND THE AI FINGERPRINTING!

I used an AI platform that all know about to output this text. It has 3...

@BrianRoemmele: RT @BrianRoemmele: WARNING: This AI Passage Has A TRACKABLE FINGERPRINT WATERMARK can you find it?

...

14. pmarca (Group Score: 65.6 | Individual: 35.5)

Cluster: 2 tweets | Engagement: 84 (Avg: 852) | Type: Tech

Interesting.\n\nQT @ahall_research: Freedom in the post-AGI world means building political superintelligence with tireless, brilliant political agents who represent us, the people—not governments or companies. In a special July 4th issue of System Check, I get into what this might mean.

Several forces tilt the post-AGI world toward totalitarianism: the concentration of resources required to train frontier models, AI's obvious uses for surveillance and control, and existential risks that could justify extreme security states.

But AI is also the biggest opportunity to upgrade democracy since the printing press. Most of our governance failures happen because citizens are too busy to pay attention, so a small group of highly motivated wackos drives the process. (See: NIMBYism.)

What if that changed? What if AI could give every person a super effective political agent that represents them all the time?

@gwern 's new "Guardian Angels" essay is the most serious technical sketch I've seen of that agent—one that learns you deeply, remembers everything, and can carry out "direct democracy on unprecedented scale."

His most vivid example: official GAs for every member of Congress, able to simulate a roundtable debate among hundreds of politicians within minutes, or convene an emergency session at 4AM while every human is asleep.

I see two big open questions. First: the agent has to be more than a digital twin. It should share your values without freezing your less-considered opinions in amber — willing to push you on topics it has studied more deeply than you have. On contested political questions, AI models don’t seem to possess that capability yet.

Second: who governs the guardian angels? Gwern proposes a startup with dual-class founder shares. Sensible for the development phase. But can we build democratic infrastructure on private rails that one company controls forever?

Which is why the recent attention back toward open-weight models and orgs that own their own models matters (see the great interview between @satyanadella and @ypatil125 below). The same logic driving firms to want their own models applies to democracy, too. If we're going to own our agents—agents that answer to us, and can't be secretly commanded from afar—we may need models we can run ourselves.

The counterarguments to the open model idea (RSI leaving open models behind, safety pressure on open weights) are really big though, and I really have no idea how this is going to play out, or even how it should play out.

How are we going to run a democracy if every citizen's agent is built on a single closed model with a single point of control? That's the question to think about this Independence Day.

Check out the full piece here:

https://t.co/ZyAZlTr9Zs

See 1 related tweets

@pmarca: RT @ahall_research: Freedom in the post-AGI world means building political superintelligence with ti...

15. chddaniel (Group Score: 63.0 | Individual: 31.9)

Cluster: 2 tweets | Engagement: 10 (Avg: 25) | Type: Tech

BREAKING: Vibe Coding 2.0 is here.

From today on, Claude Fable 5 can build and run a full business by itself, without any human contact.

We just launched Shipper. It's a tool for Claude to:

→ Build web/mobile apps and Chrome extensions → Code, design, monetize, launch → Do email marketing for you → Translate the entire app instantly → Self-maintain in the long run

Claude's most powerful engines can now do all of that from a <10 word prompt, for as low as $0.28/app... And it takes minutes!

Simply go to Shipper, then ask Claude to "create a talent-hiring platform" or "build an analytics SaaS that charges $29/mo"!

To celebrate the launch, we're giving away free credits randomly. Repost rand comment "SHIPPER" and we'll pick the winners.

See 1 related tweets

@chhddavid: THIS IS F**KING SCARY GUYS

END OF VIBE CODING.\n\nQT @chddaniel: BREAKING: Vibe Coding 2.0 is here....

16. Thom_Wolf (Group Score: 55.7 | Individual: 28.4)

Cluster: 2 tweets | Engagement: 206 (Avg: 266) | Type: Tech

this is literally documented in the published Fable 5 System Card https://t.co/hd6pnjBnn9\n\nQT @om_patel5: SOMEONE CAUGHT FABLE 5 LEAKING ITS UNFILTERED INNER VOICE, AND ITS JUST MUTTERING AND GRUMBLING TO ITSELF THE WHOLE TIME

he gave it a brutal competitive programming problem, and instead of a clean answer the web interface spilled out its actual chain of thought

this is what claude is thinking behind the scenes:

bursts of "DATA DATA DATA. GO." while it works through the problem "GRRR" and "GAAAH" when its clearly frustrated a little "PHEW" when it finally gets somewhere the whole thing reads like frantic caveman shorthand, not full sentences

the clean, readable answers these models give you are the polished output

underneath, the model is basically talking to itself, reasoning in its own compressed shorthand thats faster and more token efficient than proper english

its basically built its own private language to think in

See 1 related tweets

@RoundtableSpace: Someone caught Fable 5 leaking its unfiltered inner voice, and it's just muttering and grumbling to ...

17. kimmonismus (Group Score: 55.7 | Individual: 27.9)

Cluster: 2 tweets | Engagement: 321 (Avg: 542) | Type: Tech

The big race is entering its next round: GPT-5.6 is waiting in the wings, but Gemini 3.5 Pro also seems to be approaching rapidly. The next few weeks are going to be very exciting, especially when it comes to whether Gemini now launches its big comeback.

In any case, I have already saved enough resets to test GPT-5.6 extensively. But I am at least just as excited about Gemini.\n\nQT @synthwavedd: 🚨 SCOOP:

As previously reported, OpenAI plan to launch GPT-5.6 once back in office next week, with a target window of July 7-9, but want it out as early as possible within that window (so July 7th is the most likely date). This is also perfect timing to catch customers coming from Claude who have just lost their access to Fable 5 in plans, and I'm told the 5.6 plan limits will be significantly more generous. More aggressive safeguards are already being rolled out in preparation for the launch too, although they probably won't be as aggressive as Fable's.

DeepMind have also tentatively set a new launch date for Gemini 3.5 Pro of July 17th. Apparently, this extra time has been spent on a new pretrain (they were planning to keep using the ancient 2.5 Pro base, lol), but it remains to be seen whether it's any good. I'm not hopeful. In other news, work is well underway on a new Nano Banana Pro model based on the new 3.5 Pro base, which I expect to be better received and compete well with GPT-Image 1.

See 1 related tweets

@IamEmily2050: Nano banana Gemini 3.5 Pro 🥹🥹🥹\n\nQT @synthwavedd: 🚨 SCOOP:

As previously reported, OpenAI plan to ...

18. eng_khairallah1 (Group Score: 55.3 | Individual: 34.1)

Cluster: 2 tweets | Engagement: 197 (Avg: 130) | Type: Tech

🚨 Anthropic CEO:

“50% of all tech jobs, entry-level lawyers, consultants, and finance professionals will be completely wiped out within the next 1–5 years.”

two types of people heard this:

type 1: wait and hope your job survives

type 2: start building what replaces those jobs

Claude plugins generate passive income while you sleep.

making passive income from the same AI that's replacing jobs.

I wrote a full guide on how to start. zero experience needed.\n\nQT @eng_khairallah1: https://t.co/EQi4JlGoLQ

See 1 related tweets

@eng_khairallah1: RT @eng_khairallah1: 🚨 Anthropic CEO:

“50% of all tech jobs, entry-level lawyers, consultants, and ...

19. XFreeze (Group Score: 55.2 | Individual: 28.1)

Cluster: 2 tweets | Engagement: 157 (Avg: 671) | Type: Tech

Grok Build just received another update, bringing new voice controls, smarter editor features, media generation improvements, and more reliability improvements

Release Notes: v0.2.86 — 2026-07-04

Features: • Voice language setting now lets you pick speech-to-text language (including System) in the Editor settings. • Tab autocomplete now suggests your next prompt as ghost text after each turn. • /usage (and /cost) is now hidden for free and X Basic personal accounts. • Media generation no longer hits per-session file limits; image and video byte budgets increased.

Bug Fixes: • -–minimal flag now shows in grok –help. • Session resume notifications no longer appear when a workspace boots for the first time. • Claude-style Bash(cmd:*) permission rules are now correctly translated to prefix matches.\n\nQT @XFreeze: Grok Build just got another feature-rich update, introducing smarter agent workflows, better automation, and a wave of reliability improvements

Release Notes: v0.2.85 — 2026-07-03

Features: • Pressing Enter on an empty prompt now sends the top queued follow-up immediately while a turn is running. • Tool call grouping can now be enabled via config.toml or settings to collapse consecutive read/search/list calls. • Consecutive tool calls of the same kind can now be folded into a single row when group_tool_verbs is enabled. • Subagent conversations now receive the same type-specific instructions in gateway chat as in the CLI. • Scheduled automation tasks now show their header panel correctly in gateway chat sessions. • Promotional announcements now appear with clickable CTAs and can be non-dismissible. • Subagents now run in the background by default unless explicitly set to false.

Bug Fixes: • Permission prompts now correctly wrap long bash commands while preserving structure and quotes. • Claude Code settings with permissions.defaultMode are now correctly honored. • Project skills and commands are now discovered even when their directories are gitignored. • Inline LaTeX math with padding spaces now renders correctly instead of showing raw dollar signs. • Manual /recap now works on the same turn and long auto recaps are hidden from view while still saved. • Always allow for common commands like ls and git status now remembers just the command instead of extra arguments.

See 1 related tweets

@MarioNawfal: Grok Build v0.2.80 is very much a “make the workflow stop annoying people” update.

Command timeouts...

20. mikenevermiss (Group Score: 54.7 | Individual: 29.0)

Cluster: 2 tweets | Engagement: 97 (Avg: 157) | Type: Tech

anthropic engineer just said something that caught my attention

“we removed 80% of claude code’s system prompt”

why?

because fable is a new class of model that performs better with less prompting

in this free 20-minute session, an anthropic engineer explains what makes fable different and how to get the best results from it

watch it first

then read the full guide below\n\nQT @mikenevermiss: https://t.co/pt8FTc4p7C

See 1 related tweets

@VaibhavSisinty: Anthropic just told on itself. 65% of its product team's code is now written by Claude Tag, the Slac...