Michael Goode

@goodetrades

Penny stock trader. Short seller. Aficionado of stock fraud. AM in Cognitive Psychology. || Terms of use:

304 Following1,459 Followers6,652 Posts

Bluesky

Is this your profile?Claim it and goodetrades.dirty.social is yours.Claim profile

Michael Goode@goodetrades· 23d

Kimi comes back with a strong win. In the words of Fable Max (see image). That good review was by Kimi 3 max. Fable had found no major problems.

Evaluation: both majors are real — I verified them independently, and both escaped my review
Bottom line: this is a high-quality review. Major 1 is confirmed and is the must-fix — it's the only finding from either review that weakens a shipped safety property. Major 2 is confirmed end-to-end, including the exact test counts. The coverage-gap list is accurate on every point I checked. Most minors are accurate, a few with severity I'd adjust down because the substrate predates this commit.

Michael Goode@goodetrades· 25d

Dril Tweet: issuing correction on a previous post of mine, regarding the terror group isil. You do not, under any circumstances, gotta hand it to them

Michael Goode@goodetrades· 25d

This spec is on revision 30 and comprises take profit and stop loss orders for my trading bot. Opus 4.8 Max evaluated the reports of each LLM. Kimi found no major bugs while ChatGPT and Claude both found 2+ major bugs. Fable did the best job of finding real, severe bugs. GPT 5.6 was 2nd

Kimi found essentially zero bugs ("no normative discrepancy," everything "verified conformant")
ChatGPT found 2 High + 2 Medium + 1 Low
Claude Fable found 3 P1 + 4 P2 + spec gap + ~25 P3 (via 6 parallel sub-reviewers)

$Claude Fable 5.0 — most thorough, most real problems, by a wide margin. Method: six parallel reviewers on disjoint spec sections, coordinator re-verified every P1/P2 in code, continued the house numbering (N36+) after cross-checking the two prior reviews so nothing is re-reported. Routed every finding to an owner. Yield: all three P1s confirmed real, both sampled P2s confirmed, a confirmed spec-model gap (N52), an accurate P3 tail — zero false positives in my verified sample. Also the only report with a deep test-suite audit (e.g. it caught that the AC19 poll-shrink cell asserts coveredQty but never that the stop was resized — a real regression blind spot; SPIKE_EXIT_STOP_ADJUSTED is asserted by name in zero files). Weaknesses (real, but minor): two mislabeled call sites in N40, "per-half timestamp" in N52 is actually a single per-entry stamp, and N42's severity is generous (P2 → really P3). Correct conclusions, occasionally imperfect anchoring. It's also the longest report. ChatGPT 5.6 Sol — a genuine middle. Strengths: broadest test execution (ran 717 adjacent tests + attempted the full repo suite; independently caught the 226 SPIKE_DRY_RUN_LOCATE_INQUIRY config drift). Best operational disposition (concrete go-live checklist). And it uniquely surfaced two real items Fable didn't isolate: the cover-path empty catch at 2825 (F1) and the systemic framing of the §5 silent-catch pattern (F3, six examples vs Fable's single instance). Fatal calibration flaw: both of its headline Highs were overstated — F2 outright refuted, F1 downgraded to Low — because it repeatedly assumed promise-rejection paths that the adapter's never-reject discipline rules out (ChatGPT even conceded the adapter resolves {ok:false}, which concedes F2's premise). Both Mediums are real but hold only at Low, and F4 was already documented in the internal tester report. So it found real texture but misjudged what mattered — its top-billed bugs are its weakest. Kimi 3.0 Max — least thorough ...$

1. Claude Fable 5.0 Extra High — decisively most thorough and found by far the most real problems (8/8 verified findings confirmed, including all three P1s the internal process missed), with the best method and the only serious test-coverage audit.

2. ChatGPT 5.6 Sol Extra-High — more thorough than Kimi (it found genuine issues Kimi missed, ran the widest test execution, and caught two things even Fable didn't isolate), but hobbled by poor severity calibration — both its Highs collapsed under verification.

3. Kimi 3.0 Max — accurate but narrow; sampling-based, deferred to the tests on exactly the hard paths, and its confident "all-clear" is an unsafe false-negative.

One fairness note worth keeping separate from the ranking: on calibration/precision (no false positives, no over-rating) the order inverts — Kimi > Fable > ChatGPT. Kimi's problem isn't wrongness, it's incuriosity; ChatGPT's problem isn't laziness, it's over-claiming. Fable is the only one strong on both axes. And the three are partly complementary — the ideal review is Fable's finding set + ChatGPT's F1/F3 and test-execution breadth + Kimi's severity discipline.

Michael Goode@goodetrades· 27d

BTW, $QVCPQ and $QVCAQ will be cancelled in three weeks. Borrow is cheap on $QVCPQ so I reshorted some for that eventuality.

excerpt from spreadsheet showing 16% borrow rate on QVCPQ, 23 days to cover at zero, and potential net profit of $4897.78 on a $5k 10,000 share short position from $0.50 net of borrow fees.

Michael Goode@goodetrades· 27d

The yellow line is my first skeet about the confirmation. This is why you pay me $69,420 per year.

1 minute chart of QVCPQ today showing that the stock was trading at $5.00 at the time I skeeted that the bankruptcy plan was confirmed. Since then it dropped to under $0.20 before bouncing to $0.60.

Michael Goode@goodetrades· 27d

Amazing that some people on here are still talking about soccer and not the 20% off silk chiffon sale at Britex Fabrics

bright green blue purple plaid silk chiffon

Michael Goode@goodetrades· 27d

It was worth it.

PACER bill showing $1585.70 in charges over the last month

Michael Goode@goodetrades· 27d

PURE ALPHA

CONCLUSION
Based on the foregoing reasons, the Court hereby OVERRULES
all objections; APPROVES the Disclosure Statement; CONFIRMS the
Second Amended Plan; and RESERVES JUDGMENT on the Motion
to Terminate Exclusivity.
The Debtors are requested to submit a confirmation order
consistent with this Memorandum Decision’s Findings of Fact and
Conclusions of Law within (7) days.360
SIGNED 07/15/2026

Michael Goode@goodetrades· 28d

$LESL Leslie's (pool supplies) down 25% on the day. I think it likely that Bbg or WSJ Bankruptcy picked up the story that Octus wrote about yesterday about how they are considering bankruptcy. The tweet below was posted 2 minutes after the first LULD halt down. // Disclosure: I'm trading it

Tweet from Harvard Zhang of Octus: Leslie’s is weighing restructuring options, including a potential chapter 11 bankruptcy, as subdued demand for pool supplies and services pressures liquidity, per sources

Michael Goode@goodetrades· Jul 9

Newsbot update: of my newsbot-only strategies, I have only one strategy with a meaningful number of trades. It just hit $9k in profits with a 6.2 profit factor. Unfortunately I think this strat will max out at $50k in profit per year. I'm about to go live with strat 2.

Profit chart showing line going up and to the right from $0 to $9000 in cumulative profits

Michael Goode@goodetrades· Jul 6

haha we suck

Michael Goode@goodetrades· Jul 2

Estimating the stock stub trades at $0.20 after the final distribution makes the trade pencil out profitable but of course the borrow rate is likely to go up.

short borrow cost spreadsheet cells. APR = 400%, price shorted = $1.65, cover date = 7/30/2026, 28 days until cover, 10k shares, cover price of $0.80 (includes 0.60 final distribution which you as a short have to pay out). Borrow cost is $0.022 per day per share and net profit would be $2,277.78

Michael Goode@goodetrades· Jul 1

This is a lie. Here is a photo of minions enjoying their day off at Auschwitz

Michael Goode@goodetrades· Jul 1

Me waiting for Fable to return

Claude usage meter at 5% for 5 hour session and 7% for week ending on Saturday

Michael Goode@goodetrades· Jun 30

Still punk as fuck

Photo of Tootsie with long scar with sutures on right hind leg and hair shaved off that leg

Michael Goode@goodetrades· Jun 30

Another T12 halt (Nasdaq asking for more information) on a Chinese stock, today on $ILLR Triller. Hopefully they keep these coming. Most of these halts over the last year have lasted until the stock was delisted and many of these stocks have never traded even OTC after delisting

Daily candlestick chart of ILLR showing a big move up in the last few days from $1 to $5 and back down to $2.71

Michael Goode@goodetrades· Jun 30

"These violent delights have violent ends." So it was her own need to kill that caused her to snap her cranial cruciate ligament as she lunged for yet another bunny. Surgery went well so while she will never again be as deadly (over 100 confirmed kills) she will eventually run again.

my dog Tootsie yawning on a cheap couch in front of a bookcase

Photo of Tootsie from 3 years ago getting a bath after she killed a skunk (of course getting sprayed in the process).

Michael Goode@goodetrades· Jun 29

$ABVX Sheep wins again www.bamsec.com/filing/11931...

$LOGO Abivax Reports Positive ABTECT Maintenance Part 2 Results for Obefazimod, Demonstrating Meaningful Clinical Benefit in Refractory Ulcerative Colitis Patients and Strengthening the Phase 3 Maintenance Safety Database • Obefazimod delivered meaningful clinical benefit in a highly refractory ulcerative colitis (“UC”) population, with 37.2% of induction non-responders achieving clinical remission and 34.5% achieving endoscopic remission at Week 44 following continued 50 mg treatment • Dose escalation to obefazimod 50 mg recaptured clinical remission in 45.5% of patients who relapsed during ABTECT Maintenance Part 1, supporting a practical dose-escalation strategy for regaining and sustaining disease control over time • Across the integrated Phase 2 and Phase 3 UC program (1,704 patient-years of exposure), exposure-adjusted incidence rates (“EAIRs”) for malignancies excluding non-melanoma skin cancer (“NMSC”) were 0.35 and 0.64 events per 100 patient-years (“PYs”), and for NMSC were 0.59 and 0.64 events per 100 PYs in the all-active combined (50 mg + 25 mg) and 50 mg cohorts, respectively, all consistent with expected UC background rates • In ABTECT Maintenance Part 2, EAIRs for malignancies excluding NMSC were 0.48 and 0.69 events per 100 PYs, and for NMSC were 0.95 and 0.69 events per 100 PYs, in the all-active combined and 50 mg cohorts, respectively, all consistent with expected UC background rates • Abivax to host a conference call and webcast today at 4:30 p.m. EDT (10:30 p.m. CEST) to discuss the results PARIS, France – June 29, 2026 – 10:05 pm CEST –$

Michael Goode@goodetrades· Jun 29

Deadline for confirmation of the $QVC $QVCPQ $QVCAQ bankruptcy plan extended again to July 15th by stipulation of US Trustee and debtors. I know it was a complex, long confirmation trial but we should have a decision by now. This is why 👏 PASTA 👏 SAUCES 👏 SHOULD 👏 NOT 👏 BE 👏 JUDGES 👏

Michael Goode@goodetrades· Jun 23

The current state of LLMs

Image of a really fierce looking dragon labeled Fable, a non-existent dragon with the error API error. 500 internal server error labeled opus 4.8, a competent looking dragon labeled codex, and a deformed looking dragon labeled Gemini

Michael Goode@goodetrades· Jun 17

Some Post-Fable memes if you missed them

Red headed hillbilly from South Park at a lectern with the words "They took our Claudes!"

The one Mythos we still have, the Greek beer

Monday after Fable hits hard. Old man looking at deformed-baby-like-thing trying to put a square toy into a round hole.

Life after fable: pictures of famous Eastern-European man sitting and standing alone in various places

Michael Goode@goodetrades· Jun 17

Since Trump took Fable from us I have finally been dragged kicking and screaming into the agents / skills world. A Claude and I built a workflow that is invokable as a custom skill: /spec-pipeline to have a critic agent interact with the owner of a new spec until they reach an agreement. 1/2

Wolverine laying in bed holding an image that says "* Claude Fable 5 Mythos 5"

Michael Goode@goodetrades· Jun 17

I will stan Sheep forever but really you can't go much more wrong than following Jordan Lasker aka Cremieux, a failed grad student who makes money peddling race science on Twitter (as well as sharing a bunch of other interesting things which attracts more people to his main theme).

Tweet from Sheep of Wall Street (Biohazard3737), a >$100m net worth biotech fund manager and smart guy, saying, "Agree. Insane work by SpaceX." in response to a quote tweet of Jordan Lasker as Cremieux @cremieuxrecueil saying "Buying stuff with your equity in the middle of a huge IPO pump is an insanely good, basically perfect deal" in reference to $SPCX $SpaceX buying Cursor using stock.

Michael Goode@goodetrades· Jun 12

As planned, took the loss on that. Reshorted at 170.7 and covered half at 158.61 and half at 164.5. Paid for dinners for a week.

1 minute candlestick chart of SPCX SpaceX today showing my trades

Michael Goode@goodetrades· Jun 11

Andrew Tate and Tristan Tate are pedophiles. They had sex with 15 year old girls. Then Trump and his cronies pushed Romania to free them. PEDOPHILES. Yeah they raped and trafficked dozens of women, but the word pedophile might stick. If your brother watches/follows them? Remind him they are pedos.

When Pavel was fifteen, she told investigators, she had accompanied a school friend to a party hosted by the Tates, which was full of young girls. Her friend paired off with Tristan, leaving her with Andrew, who was so drunk that he spat as he talked. He pulled her into a bedroom and raped her, she said. Afterward, some of Tate’s “female friends” persuaded the girls to stick around for several days, and Tate kept having sex with Pavel. “We had normal, oral, and anal sexual relations, but I mention that the anal sexual acts were traumatic for me because I was in pain,” she told investigators. “Even though I asked Andrew to stop, he continued.”

Michael Goode@goodetrades· Jun 10

Stock dropped a bunch. I added 1k to my short

trade blotter from $IBKR showing selling 2k shares at 8.90 and covering 1k at 8.97 16 minutes later today

Michael Goode@goodetrades· Jun 10

I'll post this here. Obviously old. From "Senior Legal Analyst and Editor at Octus. Opinions all mine and not those of my gracious employer."

Kevin Eckhardt
@DasWreck
I’m sorry I’m a bankruptcy lawyer, the law has nothing to do with what actually happens in big bankruptcy cases. I write about bankruptcy every day and haven’t had westlaw in seven years. The company filed in Houston for a reason; they’ll get what they want.

Michael Goode@goodetrades· Jun 10

However, in coding it is like having a senior dev versus a junior dev. Fable 5 noticed 8 flaws in how I accessed SEC's EDGAR API that Opus had not noticed, and half of them were major bugs. See image. FYI on SEC filings -- don't use API if u need to extract data -- I use it for 8-K filings 2/2

It even noticed this subtle issue: "EDGAR is known to briefly 404/503 filing documents right after the ATOM entry appears. One or two quick retries (~500 ms–1 s apart) before giving up would cut these notifications and close a real missed-alert gap — a delisting filing silently skipped on a hiccup is exactly the loss mode this bot exists to avoid."

Michael Goode@goodetrades· Jun 10

Preliminary notes on Claude Fable 5: for online research with lots of resources, not much better than Opus 4.8 (they gave similar answers to this question on Starship h/t @esghound.com for the rocket equaltion knowledge). I used 1% of my Max 5x 5 hour allowance for Opus run and 8% for Fable run 1/2

$You will need to search information on the internet to answer this. I am curious about whether Elon Musk's claims for efficiency of Starship are possible. Look up information on the Tsiolkovsky rocket equation as well as efficiency (both in terms of $ per kg lifted to orbit and in terms of energy expended per kg lifted to orbit) for the Saturn V (5) rocket, Falcon X, and Falcon Heavy, and of course, SpaceX Startship to answer the question. 10:09 AM Synthesized comparative rocket efficiency analysis Synthesized comparative rocket efficiency analysis Great question — and it's one where the physics and the economics pull in opposite directions, so let me walk through both. (I'll assume "Falcon X" means the Falcon 9, SpaceX's workhorse rocket.) The Tsiolkovsky rocket equation The rocket equation says: Δv = vₑ · ln(m₀/m_f) — the velocity change a rocket can achieve equals its exhaust velocity times the natural log of the ratio of its full mass to its empty mass. Reaching low Earth orbit (LEO) requires roughly 9.3–9.5 km/s of Δv once gravity and drag losses are included, while the best chemical engines have exhaust velocities of only about 3–4.5 km/s (Raptor's specific impulse is 327 s at sea level and 380 s in vacuum, i.e. ~3.2–3.7 km/s). Because Δv needed is ~2.5–3× the exhaust velocity, the logarithm forces mass ratios of 10–20, which is why every orbital rocket is ~85–95% propellant and delivers only 2–4% of its liftoff mass as payload. Wikipedia The crucial point: the rocket equation constrains mass fractions, not dollars. It says nothing about cost. Historically, cost was dominated by throwing away the airframe and engines every flight — which is exactly the lever Musk's claims rest on. Cost per kg to LEO Here's what the numbers look like across the four vehicles: Saturn V: $185 million per launch (about $995 million in 2024 dollars) carrying 140,000 kg to LEO — roughly $7,000/kg in today's money; estimates range from about $5,000/kg to ~$8,800/kg in current …$

Starship: Musk's claim is a target of $10/kg, i.e. about $1.5 million total to deliver 150 tons to orbit, with propellant making up roughly a third of that. Crucially, that $10/kg figure is only the marginal cost of an additional flight, without paying back any development cost (development has run more than $15 billion). Independent analysts are more cautious: full reusability is expected to bring costs to roughly $1,600/kg initially, potentially falling to $100–150/kg over time, with Citi projecting ~$30/kg by 2040 in a best case but ~$300/kg if ships are only reused about 10 times. American Enterprise Institute + 3

Energy per kg to LEO
This is where it gets interesting. The theoretical minimum energy to place 1 kg in LEO (kinetic plus potential) is about 33 MJ, or roughly 9.7 kWh per kilogram — which at 100% efficiency would correspond to an energy cost of only about $1.85/kg, though rockets are not terribly energy efficient and will never approach that. NextBigFuture
Using each vehicle's propellant load and fuel heating values, the chemical energy burned per kg of payload works out approximately as follows (my calculations from published propellant masses; treat as estimates good to ±20%):

Saturn V: ~640 t of RP-1 plus ~80 t of liquid hydrogen ≈ 37 TJ; over 140 t of payload that's ~265 MJ (~75 kWh) per kg — about 12% energy efficiency.
Falcon 9: ~155 t of RP-1 ≈ 6.7 TJ; over ~17.5 t (reusable mode) that's ~380 MJ (~106 kWh) per kg — about 9% efficient.
Falcon Heavy: ~400 t of RP-1 ≈ 17 TJ; over 63.8 t that's ~270 MJ (~75 kWh) per kg — about 12% efficient.
Starship: the full stack carries methane content of approximately 1,278 metric tons (the Stanford estimate notes methane's combustion energy is about 5.57 × 10⁷ J per kg), giving ~64–71 TJ per launch; at the 150 t design payload that's ~430–470 MJ (~120–130 kWh) per kg — about 7% efficient. At 100 t it's closer to 180 kWh/kg. MobiusriskgroupStanford University
Chart showing decreasing cost per kg over time

Chart showing energy per kg of payload with Starship higher than Falcon 9 and Saturn 5

What this tells us about Musk's claim
Notice the striking pattern: the energy expended per kg has barely changed in 60 years — every chemical rocket from Saturn V to Starship burns roughly 70–130 kWh of fuel energy per kg of payload, about 7–13 times the theoretical minimum. That's the rocket equation at work. It rigidly fixes how much propellant you need, and no amount of engineering cleverness escapes it. Starship is actually slightly less energy-efficient per kg than Saturn V, partly because it hauls heavy reusability hardware (steel structure, heat shield, landing propellant) that an expendable rocket doesn't carry.
The cost, meanwhile, has varied by a factor of nearly 1,000 across vehicles burning similar energy. That's because cost was never dominated by energy — it's dominated by throwing away hardware. The propellant for a Starship launch (~4,600 tons of cheap methane and liquid oxygen) costs on the order of $900,000, with total per-flight costs of around $2 million once operations are included, according to Musk. At 150 tons of payload, $1.5M of propellant works out to roughly $7–10/kg — meaning Musk's $10/kg figure is essentially the fuel-only floor. The rocket equation permits it; nothing in physics forbids it. TechCrunch
So is the claim possible? My read:
Physically: yes. The $10/kg number doesn't violate the rocket equation or thermodynamics — it sits comfortably above the ~$2/kg pure-energy floor, and the mass ratios Starship needs are achievable with methane/LOX staged combustion.

Physically: yes. The $10/kg number doesn't violate the rocket equation or thermodynamics — it sits comfortably above the ~$2/kg pure-energy floor, and the mass ratios Starship needs are achievable with methane/LOX staged combustion.
Economically: only under heroic assumptions. It requires full and rapid reusability (boosters and ships reflown hundreds of times with airline-style turnaround), enormous flight rates so fixed costs amortize away, near-zero refurbishment, and — as the NASA analysis notes — it ignores repaying the development cost entirely. Reality is still catching up: current Starship versions have flown with actual demonstrated LEO payloads of 15–35 tons for Blocks 1–2, with 100 tons targeted for Block 3, and a record of 7 successes in 12 flights, and most analysts consider $100–500/kg more realistic in the near-to-medium term. WikipediaSpaceNexus
Even the skeptics' number is remarkable, though: $100–300/kg would be 5–15× cheaper than Falcon Heavy and 20–70× cheaper than Saturn V in real terms. The honest summary is that Musk's $10/kg is an aspirational marginal-cost limit — the price of fuel for a fully amortized, airline-like system — not a near-term ticket price. The rocket equation guarantees Starship will never be energy-efficient; it just doesn't have to be, because energy was never the expensive part.

Michael Goode@goodetrades· Jun 9

Happy delivery day for Jack Schwager's "Market Wizards: The Next Generation." I personally know three of the traders in the book: Lukas Froelich, Phil Goedeker, and Lance Breitstein. They are all amazing traders and Lukas will someday be ranked as one of the greatest traders of all time.