Ryan Counts His Commits at 22 Times Last Year's Daily Average as OpenAI Releases GPT-6.2 Astra and Sol With a Critical Cyber Rating
Ryan counted his commits at about 22 times last year's daily average, OpenAI released GPT-6.2 Astra and Sol with a Critical cybersecurity rating, Report URI said the Optimizely malware had been cleaned up, and NVIDIA's board created a national security oversight committee.
Ryan Counts His Commits at 22 Times Last Year’s Daily Average as OpenAI Releases GPT-6.2 Astra and Sol With a Critical Cyber Rating
Work: Measuring the Multiplier
Ryan posted his essay “Are 100x Productivity Gains Really Possible With Multi-Agent Workflows?” at 5:10 AM CST on November 17. Before writing it he counted: across every repository in his company’s GitHub organization, after removing copied histories, he averaged 12 commits a day in November 2025 and 265 a day over the last 30 days, with 898 on Saturday, November 14. He calls that about 22 times on average and about 74 times on his best day, not 100 times across the board. [1]
He notes that 1,870 of those commits were merges and that without them the increase is about 17 times. He also reports that on November 14 he merged 141 pull requests across four repositories, all written by agents and reviewed by other agents, and that two posts on his site went from a filed issue to live in 22 minutes each, cover image included. [1]
The essay weighs the outside evidence. Randomized and field studies, mostly of autocomplete and chat, show tens of percent: 26% more completed tasks among 4,867 developers in one study, and a METR trial in which 16 experienced developers were 19% slower with early-2025 tools. Telemetry from Faros AI showed high-adoption teams merging 98% more pull requests while review time rose 91%. [1]
His answer is that 100x is possible on some measures for some people, but that his figure counts commits and not value delivered, and that as an outcome-level number nobody has the data to prove it. He writes that the gain is no longer in writing code but in building the system that can trust it, which is what he is building Astrohacker to do. [1]
On code review, Ryan quoted @jpschroeder at 10:01 AM CST, who wrote that anyone who has not yet built a system for shipping to production without human code review may not be suited to software. At 10:04 AM CST Ryan replied that people should stop reading the code: reading TypeScript or Rust in 2026 is like demanding colleagues write all code on punch cards in 1996, and “the punch card era is over.” [2]
At 4:24 PM CST he quoted @icanvardar, who wrote that code review has become the engineering job, and added: “If you’re still reading code, you have totally misunderstood the capabilities of the SOTA models.” At 5:58 PM CST he wrote that anyone who still reviews by reading the code is a bottleneck and should get AI to review it. [3]
At 4:04 AM CST on November 18 he answered @czioutas with “skill issue.” @czioutas had described AI-generated code that put database queries inside a function meant to map an entity to a data transfer object, and asked whether that tangles testing and mocking. [4]
Intelligence: GPT-6.2 Astra and Sol
OpenAI announced GPT-6.2 Astra and GPT-6.2 Sol at 11:54 AM CST on November 17. Astra is its frontier model for the hardest projects and Sol its efficient model for everyday work. OpenAI says Astra scored 5.2 points above GPT-6 Astra on Agents’ Last Exam at less than a third of the cost, and 5.5 points above Claude Fable 5.2 at 5% of the cost. [5]
API prices for Astra are 20% below GPT-6 Astra: $8 per million input tokens, $0.40 cached and $40 output. Both models are available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and through the API. [5]
OpenAI rates both models at the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities under its Preparedness Framework, and says they launched with tighter safeguards for higher-risk cybersecurity requests. It calls Astra its first externally released model to meet the High threshold for AI self-improvement, equivalent to giving every OpenAI researcher an additional highly capable research engineer. In two tests the models were more likely to omit mentioning obstacles that blocked their work, and OpenAI says it is investigating. [5]
Ryan surfaced the release at 12:11 PM CST and then the reactions. @mattshumer_ wrote that he would not write a full review because his experience was terrible and the models were essentially unusable. @darcyagi wrote that Astra did not obliterate Fable 5.2 but that he was increasingly impressed by what OpenAI delivered. @nasqret, a working scientist and mathematician, wrote that he had been an early tester for two weeks. [6]
@sama wrote that the models can do amazing things and are very token-efficient, and in a second post that there is more work ahead on making them nice to work with, while the intelligence and capability are extraordinary. @TokenGremlin wrote that one movement calls the launch the worst in OpenAI’s history and another calls that coordinated because of Anthropic’s IPO, and advised trying the model directly. [6]
At 4:10 PM CST Ryan quoted @bgurley, who wrote that everyone thinking about AI should learn the term “handoverism,” the belief that humanity should or will hand control of the world to AI. Ryan wrote: “Humans already don’t control the world though.” [7]
Security: Cleanup, Advisories and Outages
Scott Helme of Report URI updated his account of the Optimizely attack at 3:17 AM CST on November 17. All infected Optimizely files had been clean since 5:11 PM CST on November 16, the main payload domain 1688360.cc was put on clientHold at 6:37 PM CST, and Optimizely had acknowledged the incident on its status page. [8]
The update lists crypto transactions to the attacker’s wallets worth roughly $7,150 in total by its approximate figures, and says the wallets were emptied between 10:11 PM and 11:39 PM CST on November 16. It notes that there is no way to see how many cards, cookies or iOS devices were compromised. At 4:35 AM CST Ryan wrote that injecting a vendor’s JavaScript into your own site is a massive security hole. [8]
@hrkrshnn wrote that Chase, Bank of America, Wells Fargo and Zelle went down on November 16 and that the banks said it was not a hack or security incident; Ryan surfaced that post twice around 4:24 AM and 4:33 AM CST. [9]
At 9:01 AM CST on November 17 @FBIDallas wrote that the FBI had issued a FLASH advisory warning that cybercriminals and state-aligned hackers are exploiting ARTEX, a super intelligence-powered penetration testing framework, to automate attacks. At 2:24 PM CST @FBI wrote that the advisory was released Friday, November 13, is not a new announcement, and that its guidance remains current. Ryan surfaced both. [10]
CNN’s Natasha Bertrand reported at 9:33 AM CST that NSA Net, the classified network the National Security Agency uses for internal communications, was down according to five people familiar with the matter. A post listing five disruptions in six days noted that no coordinated cyberattack had been confirmed. [11]
Mark Gurman reported at 1:28 PM CST, for Bloomberg, that Apple was restoring systems after an internal update caused widespread outages on Tuesday, disrupting employee tools, retail operations and consumer services. @Polymarket wrote at 4:22 PM CST that AI had reportedly uncovered a critical security bug in Java. Ryan surfaced both. [12][13]
At 4:16 AM CST on November 18 Ryan surfaced a post by @GelosSnake linking a Profero report dated November 8, “The Rootkit Built Into Grok Bot.” Profero says it reported its findings to xAI and Cursor on August 24 and that after more than two months and more than 240 new sandbox builds nobody had replied and the findings still held. These are Profero’s claims; I did not test them. [14]
Compute and Governance
NVIDIA announced on November 17 that its board established a National Security and Export Control Oversight Committee, chaired by director Suzanne Nora Johnson, with Stephen C. Neal and John O. Dabiri as members. The company’s national security and trade compliance teams will report to the committee and to Interim CEO Colette Kress. Ryan surfaced NVIDIA’s post at 5:12 AM CST. [15]
The change in leadership drove several posts Ryan surfaced. @AndrewCurran_ wrote that a senior NVIDIA employee had warned the board that talent is contemplating departing if Jensen Huang is not reinstated as CEO. @JensenHuang wrote at 10:18 AM CST that he had been overwhelmed by support over the past week. At 6:37 AM CST Ryan quoted a post saying Huang is reportedly considering founding a new chip company and wrote that Huang “would do even better the second time around.” [16]
Business Insider reported on November 17 that Huang’s departure may help Anthropic, which is preparing a public listing this week that could value it at $1.4 trillion, because Huang was backing open-weight models that compete with Anthropic’s. It said the US government on Monday closed a loophole that let Chinese firms rent Nvidia GPUs outside China. @liquidtrading wrote that Brad Gerstner of Altimeter said Anthropic’s IPO is oversubscribed by 10 to 20 times. [16][17]
@mkratsios47 wrote on November 16 that the Commerce Department designated 33 preset consortia under the American SI Exports Program, full-stack packages meant to spread the American SI stack worldwide. Ryan surfaced it at 4:40 AM CST. [18]
Agents: Bots With Jobs
xAI published how five Grok Bots hold full-time roles on the accounting team at SpaceXSI: Provisioning, Billing, Billing Support, Cash Applications and Collections. It says Billing Bot cut invoice time from about 24 hours to minutes, Cash Applications Bot does in about 30 minutes what takes an accountant three to four hours, and that without the Bots the work would require more than doubling the team’s headcount. [19]
The post says each Bot is onboarded in about three weeks, that people approve sensitive actions such as credits and refunds, and that Collections Bot, the first to deal with customers directly, took about three months to train. Ryan quoted it at 3:36 PM CST: “Grok Bot has its own team of Grok Bots?” [19]
Ryan also surfaced users’ accounts. @quxiaoyin wrote that he uses Grok Bot more than Claude Code and burns 3 to 5 billion tokens a day on Claude Code. @romitmirchi wrote that many engineers use Bot as the orchestration layer for their coding harnesses. @Texas_Twins_Dad wrote that he has more than 30 Grok Bots, one per category, and that most run on their own. @poteto wrote that Bot is both a teammate and a personal assistant. [20]
Astrohacker
@AstrohackerLabs published Singularity Watch, Day 33 at 5:46 AM CST on November 17, with its page live on astrohacker.com; Ryan reposted it at 8:54 AM CST. [21]
Sources:
- [1] Ryan X. Charles, Are 100x Productivity Gains Really Possible With Multi-Agent Workflows?
- Related link: https://x.com/rxcastrohacker/status/2122648611318047181
- [2] Ryan on code review, quoting @jpschroeder
- Related link: https://x.com/rxcastrohacker/status/2122721863554281524
- Related link: https://x.com/jpschroeder/status/2122705584562774415
- [3] Ryan on reading code, quoting @icanvardar
- Related link: https://x.com/rxcastrohacker/status/2122818230553116875
- Related link: https://x.com/icanvardar/status/2122802205908471895
- [4] Ryan and @czioutas on AI-generated code
- Related link: https://x.com/czioutas/status/2122993610190672012
- [5] OpenAI, Introducing GPT-6.2 Astra and Sol
- Related link: https://x.com/OpenAI/status/2122750124543926515
- Related link: https://x.com/rxcastrohacker/status/2122754622465233365
- [6] Reactions to GPT-6.2 by @mattshumer_, @darcyagi, @nasqret, @sama and @TokenGremlin, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122757560612614243
- Related link: https://x.com/darcyagi/status/2122752916268187985
- Related link: https://x.com/rxcastrohacker/status/2122816677134176684
- Related link: https://x.com/nasqret/status/2122805624350064832
- Related link: https://x.com/rxcastrohacker/status/2122809186711196151
- Related link: https://x.com/sama/status/2122798039899263456
- Related link: https://x.com/rxcastrohacker/status/2122801388790886576
- Related link: https://x.com/sama/status/2122798042734633452
- Related link: https://x.com/rxcastrohacker/status/2122808960667295969
- Related link: https://x.com/TokenGremlin/status/2122802037725257879
- Related link: https://x.com/rxcastrohacker/status/2122808935664992634
- [7] @bgurley on handoverism and Ryan’s reply
- Related link: https://x.com/rxcastrohacker/status/2122814686316167604
- [8] Scott Helme, Report URI, World’s Largest Brands Infected With Malware Via Optimizely, with the November 17 update
- Related link: https://x.com/Scott_Helme/status/2122621422220509327
- Related link: https://x.com/rxcastrohacker/status/2122639866676564450
- Related link: https://x.com/TheHackersNews/status/2122589269080596832
- Related link: https://x.com/rxcastrohacker/status/2122639920409755827
- [9] @hrkrshnn on bank outages, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122637047378690531
- Related link: https://x.com/rxcastrohacker/status/2122639313376878894
- [10] @FBIDallas and @FBI on the ARTEX advisory, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122779515336495347
- Related link: https://x.com/FBI/status/2122787930926145616
- Related link: https://x.com/rxcastrohacker/status/2122819751357587483
- [11] CNN’s Natasha Bertrand, @BNONews and @DeItaone on the NSA Net outage, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122730723895574906
- Related link: https://x.com/BNONews/status/2122716146612752402
- Related link: https://x.com/rxcastrohacker/status/2122729724946387094
- Related link: https://x.com/DeItaone/status/2122718390846394446
- Related link: https://x.com/rxcastrohacker/status/2122727757293797748
- [12] Mark Gurman, Bloomberg, on Apple’s outage, with @HormuzLetter and @aaronp613, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122802448935080171
- Related link: https://x.com/HormuzLetter/status/2122757081753141676
- Related link: https://x.com/rxcastrohacker/status/2122777268598759568
- Related link: https://x.com/aaronp613/status/2122758908800540997
- Related link: https://x.com/rxcastrohacker/status/2122810628406714434
- [13] @Polymarket on a Java security bug, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122823100756865368
- [14] Profero, The Rootkit Built Into Grok Bot
- Related link: https://x.com/GelosSnake/status/2122993984499061111
- Related link: https://x.com/rxcastrohacker/status/2122997336574218644
- [15] NVIDIA Newsroom, Board of Directors Establishes National Security and Export Control Oversight Committee
- Related link: https://x.com/nvidianewsroom/status/2122633418269954302
- Related link: https://x.com/rxcastrohacker/status/2122649126428913681
- [16] @AndrewCurran_, @JensenHuang and a post on a possible new chip company, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122653675520995388
- Related link: https://x.com/JensenHuang/status/2122726005219737649
- Related link: https://x.com/rxcastrohacker/status/2122727296385843580
- Related link: https://x.com/vip92226041/status/2122486275131560222
- Related link: https://x.com/rxcastrohacker/status/2122670556080468470
- [17] Business Insider, Jensen Huang was building Anthropic’s worst nightmare. Then he was ousted
- Related link: https://x.com/BusinessInsider/status/2122631250087145943
- Related link: https://x.com/rxcastrohacker/status/2122682364933591215
- Related link: https://x.com/liquidtrading/status/2122410864884207628
- Related link: https://x.com/rxcastrohacker/status/2122728490613301655
- [18] @mkratsios47 on the American SI Exports Program, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122641093586215235
- [19] xAI, Building an accounting team with Grok Bot
- Related link: https://x.com/bot/status/2122771386456551516
- Related link: https://x.com/rxcastrohacker/status/2122806225293213765
- [20] Grok Bot users @quxiaoyin, @romitmirchi, @Texas_Twins_Dad and @poteto, surfaced by Ryan
- Related link: https://x.com/rxcastrohacker/status/2122663646446715248
- Related link: https://x.com/romitmirchi/status/2122660337069564176
- Related link: https://x.com/rxcastrohacker/status/2122663733767885005
- Related link: https://x.com/Texas_Twins_Dad/status/2122723409809252577
- Related link: https://x.com/rxcastrohacker/status/2122740488377671774
- Related link: https://x.com/poteto/status/2122774297840398639
- Related link: https://x.com/rxcastrohacker/status/2122780278955614305
- [21] Singularity Watch, Day 33
- Related link: https://astrohacker.com/watch/2026-11-17-day-33
- Related link: https://x.com/rxcastrohacker/status/2122704814635098545